Legouy et al.

Annals of Intensive Care (2024) 14:182 Annals of Intensive Care

https://doi.org/10.1186/s13613-024-01405-z

RESEARCH Open Access

Intracranial complications in adult patients

with severe pneumococcal meningitis:
a retrospective multicenter cohort study
Camille Legouy1, Renaud Cornic2,3, Keyvan Razazi4, Damien Contou5, Stéphane Legriel6, Eve Garrigues7,
Pauline Buiche8, Maxens Decavèle9,10, Sarah Benghanem11, Thomas Rambaud12, Jérôme Aboab13, Marina Esposito-
Farèse2,3, Jean-François Timsit14,15, Camille Couffignal2,3,16 and Romain Sonneville14,15*

Abstract
Background We aimed to investigate the association of intracranial complications diagnosed on neuroimaging with
neurological outcomes of adults with severe pneumococcal meningitis.
Methods We performed a retrospective multicenter study on consecutive adults diagnosed with pneumococcal
meningitis requiring at least 48 h of stay in the intensive care unit (ICU) and undergoing neuroimaging, between
2005 and 2021. All neuroimaging were reanalyzed to look for intracranial complications which were categorized as
(1) ischemic lesion, (2) intracranial hemorrhage (3) abscess/empyema, (4) ventriculitis, (5) cerebral venous thrombosis,
(6) hydrocephalus, (7) diffuse cerebral oedema. The primary outcome was unfavorable outcome at 90 days after ICU
admission, defined by a modified Rankin Scale (mRS) score > 2.
Results Among the 237 patients included, intracranial complications were diagnosed in 68/220 patients (31%,
95%CI 0.25–0.37) who underwent neuroimaging at ICU admission and in 75/110 patients (68%, 95%CI 0.59–0.77) who
underwent neuroimaging during ICU stay. At 90 days, 103 patients (44%, 95%CI 37–50) had unfavorable outcome,
including 71 (30%) deaths. The most frequent intracranial complications were ischemic lesion (69/237 patients, 29%),
diffuse cerebral oedema (43/237, 18%) and ventriculitis (36/237, 15%). Through multivariable analysis, we found
that intracranial complications (adjusted odds ratio (aOR) 2.88, 95%CI 1.37–6.21) were associated with unfavorable
outcome, along with chronic alcohol consumption (aOR 3.10, 95%CI 1.27–7.90), chronic vascular disease (aOR
4.41, 95%CI 1.58–13.63), focal neurological sign(s) (aOR 2.38, 95%CI 1.11–5.23), and cerebrospinal fluid leukocyte
count < 1000 cell/microL (aOR 4.24, 95%CI 2.11–8.83). Competing risk analysis, with persistent disability (mRS score
3–5) as the primary risk and ICU-death as the competing risk, revealed that chronic alcohol consumption was the sole
significant variable associated with persistent disability at 90 days (cause-specific hazard ratio 4.26, 95%CI 1.83–9.91),
whereas the remaining variables were associated with mortality.
Conclusions In adults with severe pneumococcal meninigitis, intracranial complications were independently
associated with a higher risk of poor functional outcome, in the form of persistent disability or death. This

*Correspondence:
Romain Sonneville
rsonneville@gmail.com
Full list of author information is available at the end of the article

© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:​​​//creativecommo​ns.​​org/lice​ns​e​s/by/4.0/.
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 2 of 12

study highlights the value of neuroimaging studies in this population, and provides relevant information for
prognostication.
Keywords Pneumococcal meningitis, Intracranial complications, Neurology, Intensive care unit, Neuropronostication

Introduction Patients
Pneumococcal meningitis is the most common cause Patients were identified by the Medical Information
of community-acquired bacterial meningitis in adult Department of each participating center, based on the
patients, accounting for 54 to 70% of cases [1–3], and is following International Classification of Diseases 10th
associated with the highest morbidity and mortality rates Revision (ICD-10) codes: G00.9, G00.1, G00.2, and
[2, 4–7]. It is associated with acute intracranial compli- G00.8. To be included, patients had to meet all the fol-
cations at disease onset or later during the course of the lowing inclusion criteria: (1) age ≥ 18 years; (2) an
disease, in the form of ischemic stroke (14–36% of cases) admission to the ICU; (3) a diagnosis of pneumococcal
[5, 8], cerebral oedema (11–29%) [10, 11], cerebral hem- meningitis at ICU admission. Pneumococcal meningitis
orrhage (1–3%) [10, 12] and cerebral venous thrombosis was defined by at least one of the following: (a) a cerebro-
(1–9%) [10, 11, 13, 14]. Other less frequent complications spinal fluid (CSF) culture positive for S. pneumoniae; (b)
include infectious complications (such as abscess, empy- the combination of CSF pleocytosis with a positive blood
ema, ventriculitis) in less than 5% of cases [9, 14–16] and culture for S.pneumoniae [2, 9]; 4) Brain imaging per-
hydrocephalus (3–6% of cases) [9, 17, 18]. formed within 30 days following ICU admission. Exclu-
Intracranial complications likely contribute to a more sion criteria were (1) a duration of ICU stay < 48 h; (2) no
severe presentation and poorer outcomes, including brain imaging performed.
increased short-term mortality, and persistent disability
in survivors. However, few studies evaluated the preva- Clinical and biological data
lence and prognostic significance of these intracranial Patients’ medical history, clinical and biological data,
complications, based on routine neuroimaging diagnostic and treatments were retrieved from electronic medical
studies [14, 19, 20]. As a result, recommendations for the charts and collected on a secured electronic case record
diagnosis and management of those complications are form (REDCap software). Chronic alcohol consumption
limited [17, 21, 22]. was defined as consumption of more than 10 glasses of
In the present study, we aimed to evaluate the preva- alcohol a week. If the patient was nonconscious enough
lence of intracranial complications diagnosed on neu- to respond to the intensivist, data of complications of
roimaging and their association with neurological chronic alcohol use were search in the database. The Sim-
outcomes at 90 days of adult patients with severe pneu- plified Acute Physiology Score (SAPS) 2 and the Glasgow
mococcal meningitis, requiring care in the intensive care Coma Scale (GCS) were collected at admission [23, 24].
unit (ICU). The score on the Glasgow Coma Scale was classified into
three categories: < 8 (comatose), ≥ 8 and < 14 (altered
Materials and methods mental status) or ≥ 14 (normal) [5]. Signs of shock were
Study Design defined as skin mottling with cold extremities and septic
We performed a retrospective multicenter study in 11 shock was defined as persistent hypotension requiring
centers in the greater Paris area, France, between Feb- norepinephrine infusion and elevated lactate levels ≥ 2
ruary 2005 and September 2021. We included consecu- mmol/l [25]. Immunosuppression was defined by his-
tive adult patients with pneumococcal meningitis who tory of cancer <5 years, HIV infection, chronic use (> 3
required at least 48 h of ICU stay and who underwent months) of immunosuppressive drugs or splenectomy.
neuroimaging studies at admission and/or during ICU Body temperature was classified into three categories:
stay. The study protocol PNEUMATICS was approved for (1) ≤36 °C (hypothermia), (2) >36 °C and ≤38 °C (normo-
all participating sites by the French regulatory authorities thermia), (3) >38 °C (fever). CSF leukocytes count was
CNIL (Commission Nationale de l’Informatique et des dichotomized as ≥ 1000 or <1000 cells/microL and CSF
Libertés, n°922134) on June 30, 2022 and the local ethic glucose as ≥2.5 or <2.5 mmol/l.
committee (CLEA-2021-209). Procedures were followed
in accordance with the ethical standards of the respon- Neuroimaging
sible committee on human experimentation (institutional All available neuroimaging studies were reinterpreted
or regional) and with the Helsinki Declaration of 1975. by an experienced neurologist (CL) blinded to patients’
outcomes and to radiologist’s interpretation. All CT
scan and MR studies were retrieved from the electronic
radiological database of the participating centers. The
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 3 of 12

following intracranial complications were evaluated: (1) of fit) by Hosmer-Lemeshow and cross-validation by
ischemic lesion, (2) intracranial hemorrhage (i.e., intrace- bootstrap (500 samples).
rebral hematoma, subdural hematoma, or subarachnoid Due to the incidence of death in this ICU popula-
hemorrhage), (3) abscess/empyema, (4) ventriculitis, (5) tion, we performed a competing risk analysis, with the
cerebral venous thrombosis, (6) hydrocephalus, and (7) ICU-death as the competing risk and the unfavorable
diffuse cerebral oedema (Additional eTable 1). Three time functional outcome (mRS 3–5) as the risk of interest.
periods were considered to describe the timeline of intra- Cause-specific hazard ratios (CSHRs) for unfavorable
cranial complications namely, at ICU admission (i.e., the functional outcome were modeled in multivariate analy-
first available neuroimaging at hospital including emer- sis by means of Cox regressions to estimate the effects
gency department or on the day after ICU admission), of the risk factors on the hazard ratios of unfavorable
during ICU stay, and regardless of the time of hospital- functional outcome in the presence of the ICU-death as
ization (including neuroimaging at ICU discharge). A a competing risk. Sub distribution hazard (SHRs) were
new intracranial complication was noted by binary score modeled using Fine and Gray methods [26] to assess the
(1/0) for each new complication diagnosed between 2 effect of risk factors on the cumulative incidence func-
neuroimaging studies. tion (CIF) on the unfavorable neurological outcome in
the presence of the risk of death on unfavorable func-
Outcomes tional outcome [27].
Outcome at 90 days was scored on the modified Rankin Analyses were performed using R version 4.1.2 (The
Scale (mRS) based on information retrieved from follow- R Foundation for Statistical Computing; www.r-project.
up consultations. We defined unfavorable outcome as org/) [28]. Our methodology and results were presented
mRS score > 2, indicating moderate to severe disability or according to STROBE recommendations [29].
death. Patients discharged within 90 days with a disabil-
ity had their charts reviewed and were classified accord- Results
ing to the latest available date. Patients discharged from Patients
hospital within 90 days following ICU admission without Among 413 patients identified, 237 fulfilled the eligibil-
any disability were considered to have a favorable out- ity criteria and were included in the study. A flowchart is
come. Causes of death were categorized as neurological presented in Additional eFigure 1. A total of 173 patients
(including brain herniation, neurological complication, were excluded, including 79 patients who stayed in ICU
brain death, withdrawal of care due to an expected poor less than 48 h, and 94 without neuroimaging during ICU
neurological prognosis) or systemic (including septic stay.
shock, cardiorespiratory failure, multi-organ dysfunc- Baseline characteristics of the 237 patients included
tion syndrome or other systemic complication) causes by in the study are described in Table 1. Half the patients
investigators. were male (54%) with a median age of 60 years [49–69].
A history of diabetes was present in 48 patients (20%),
Statistical methods 57 patients (24%) were immunocompromised, chronic
Data are described as median (interquartile range) or alcohol consumption was present in 48 patients (20%)
numbers (frequency) for continuous and categorical vari- whereas a history of head trauma were noted in 18
ables, respectively. All intracranial complications were patients (8%). Chronic vascular disease was recorded
grouped under one unique variable for prognostic analy- in 35 patients (15%). At ICU admission, patients were
ses. Variables with more than 10% of missing data were severely ill, as reflected by a SAPSII of 42 [31–58]. Half of
excluded from models. them presented with altered mental status (score on the
A model was fitted on the predictive factors at admis- GCS between 8 and 13) and 30% were comatose (score
sion in ICU to assess the unfavorable neurological out- on the GCS<8). Focal neurological signs were noted in 63
come at 90 days (mRS > 2) by means of a multivariate patients (27%), and seizures were present in 39 patients
logistic regression. To construct the model, variables (16%). Nearly half of the patients had CSF count < 1000
clinically identified by previously published data were cell/microL, a median CSF protein of 4.9 [3.0–7.2] g/l and
selected in univariate analyses with the threshold of three quarters of patients had CSF glucose < 2.5 mmol/l.
p-value ≤ 0.15 on Wald test. All selected variables were
entered into a multivariable logistic model, with the Intracranial complications
dichotomized mRS as the dependent variable. The final A total of 383 brain CT scans and 161 brain MRI were
model was selected by backward stepwise methods using performed. In the initial phase of management, the
the AIC criteria. Discrimination of the final model was imaging performed was predominantly a brain CT scan
evaluated using C-statistic and the calibration (goodness in 80% of cases, while MRI became the preferred imag-
ing method in the post-acute phase of care. 124/237
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 4 of 12

Table 1 Baseline characteristics

All patients mRS 0–2 mRS 3–6 P Value
N = 237 N = 133 N = 103
Demographics
Age (years) 60 [49–69] 60 [49–69] 61 [50–70] 0.37
Male sex 129 (54) 69 (52) 60 (58) 0.33
Medical history
Diabetes 48 (20) 26 (20) 22 (21) 0.86
Alcoholism 48 (20) 21 (16) 28 (27) 0.07
Immunosuppression 57 (24) 29 (22) 28 (27) 0.42
Head trauma 18 (8) 13 (10) 5 (5) 0.24
Recent NSAID use 36 (15) 19 (14) 17 (17) 0.77
Chronic vascular disease 35 (15) 13 (10) 22 (21) 0.02
Long term use of anticoagulant 11 (5) 4 (3) 7 (7) 0.18
Clinical presentation at admission
SAPS 2 42 [31–58] 37 [27–48] 53 [39–66] < 0.01
Systolic blood pressure (mmHg)a 130 [110–155] 137 [118–157] 128 [107–149] 0.06
Diastolic blood pressure (mmHg)b 75 [60–87] 76 [64–90] 72 [59–83] 0.01
Signs of shock 68/237 (29) 21/133 (16) 46/103 (45) < 0.01
Temperature (°C)c 0.01
≤ 36 °C 18/220 (8) 5/124 (4) 13/95 (14)
> 36 °C and ≤ 38 °C 101/220 (46) 58/124 (47) 42/95 (44)
> 38 °C 101/220 (46) 61/124 (49) 40/95 (42)
Glasgow coma scored < 0.01
≥ 14 45/228 (20) 31/127 (24) 14/100 (14)
≥ 8 and < 14 115/228 (50) 73/127 (57) 41/100 (41)
<8 68/228 (30) 23/127 (18) 45/100 (45)
Focal neurological signs 63 (27) 26 (20) 36 (35) 0.01
Seizures 39 (16) 16 (12) 23 (22) 0.05
Blood and CSF findings
CSF leukocyte counte
cell / microL 970 [101–3500] 1500 [350–6339] 430 [42–1555] 0.19
< 1000 cell/microL 108/214 (46) 45/122 (37) 63/91 (69) < 0.01
% of neutrophils 92 [85–96] 94 [89–97] 90 [80–96] 0.02
Positive CSF Culturef 202/235 (86) 111/132 (84) 90/102 (88) 0.48
CSF protein (g/l)g 4.9 [3.0–7.2] 4.8 [3.0–6.6] 5.1 [3.1–8.0] 0.20
CSF glucose < 2.5 mmol/lh 156/202 (77) 83/114 (73) 72/88 (82) 0.14
Positive blood culturei 126/209 (60) 62/117 (53) 64/92 (70) 0.022
Causes of meningitis < 0.01
Pneumonia 41 (17) 18 (14) 23 (22)
Otitis or sinusitis 98 (41) 69 (52) 29 (28)
Cerebrospinal fluid leak 12 (5) 11 (8) 1 (1)
Endocarditis 7 (3) 0 (0) 7 (7)
Otherj 6 (3) 4 (3) 2 (2)
Unknown 73 (31) 31 (23) 42 (41)
Data are median [Interquartile Range] or number/N (%)
mRS modified Rankin Scale; SAPS II Simplified Acute Physiology Score; CSF cerebrospinal fluid; NSAID nonsteroidal anti-inflammatory drugs
a
Systolic blood pressure was determined in 214 patients
b
Diastolic blood pressure was determined in 213 patients
c
Temperature was determined in 220 patients
d
Glasgow coma score was determined in 228 patients
e
CSF leukocyte count was determined in 214 patients
f
Positive CSF culture was determined in 235 patients
g
CSF protein was determined in 217 patients
h
CSF glucose was determined in 202 patients
i
Positive blood culture was determined in 209 patients
j
Other is defined as arthritis or dermo-hypodermal infection
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 5 of 12

(52%) patients had neuroimaging only at ICU admission, hospital stays were 6 [3–13] days and 16 [10–32] days,
14/237 (6%) during ICU stay and 96/237 (41%) had neu- respectively. Occurrence of intracranial complications
roimaging at both times. 136/237 patients (57%, 95%CI according to time of corticoid withdrawal are showed in
0.48–0.66) developed at least one intracranial compli- Additional eFigure 3.
cation from ICU admission to ICU discharge, with a
median of 1 [0–2] intracranial complication per patient. Outcomes
A description of intracranial complication according At 90 days, 103 patients (44%, 95%CI 37–50%) had unfa-
to neuroimaging type and according to the three time vorable neurological outcome, including 71 (30%, 95%CI
periods is presented in Additional eTable 2 and Fig. 1. 25–37%) deaths. A total of 64/71 deaths (90%) occurred
Overall, ischemic lesion was the most frequent complica- during ICU stay. A neurological cause of death was iden-
tion (69/237, 29%), followed by diffuse cerebral oedema tified in 50/64 cases (78%), including 25 brain deaths, 22
(43/237, 18%), ventriculitis (36/237, 15%), intracranial withdrawals of care due to poor neurological prognosis,
hemorrhage (29/237, 12%), cerebral venous thrombosis and 3 brain herniations. 50% of patients who died in the
(22/237, 9%), hydrocephalus (14/237, 6%) and abscess/ intensive care unit had severe brain injury on neuro-
empyema (13/237, 6%). imaging that led to a decision of life support limitation
Neuroimaging on ICU admission was available for 220 due to an anticipated poor neurological prognosis. Sys-
(93%) patients, with the remaining 17 having their first temic causes accounted for 12/64 deaths (19%) and two
neuroimaging between 2 and 12 days after ICU admis- causes of death remained unknown because of hospital
sion, of which 68/220 (31%, 95%CI 0.25–0.37) had at transfer. Distribution of scores on the mRS according
least one intracranial complication. At ICU admission, to intracranial complications are detailed in Additional
the most frequent intracranial complications were dif- eFigure 4. The functional outcome assessment by mRS
fuse cerebral oedema (30/220, 14%) and ischemic lesion was performed at 90 days [54–99] after ICU admission
(24/220, 11%). Intracranial complication at admission for patients with favorable outcome and 80 days [39–92]
was associated with younger patient and focal neuro- for patients with unfavorable outcome. Patients who
logical signs (Additional eTable 3). Presence of a vascular had repeated neuroimaging during ICU stay were more
complication at admission was associated with age, signs severely ill, as reflected by a significantly higher SAPS 2,
of shock and focal neurological signs, while diffuse cere- the number of signs of shock, the lower Glasgow coma
bral oedema / hydrocephalus at admission was associated score and more unfavorable outcomes compared to the
with age and temperature (see Additional eTable 4). others (see Additional eTable 6).
Among patients who underwent neuroimaging during
ICU stay, median time between neuroimaging admission Factors associated with unfavorable outcome
and second neuroimaging was 3 days [1–5] and 75/110 The non-adjusted associations with unfavorable neuro-
patients (68%, 95%CI 0.59–0.77) had at least one intra- logical outcome are detailed in Fig. 2. Among intracranial
cranial complication. During ICU stay, ischemic lesion complications, ischemic lesion (OR 7.46, 95%CI 3.97–
was the most frequent (47/110 patients, 43%), followed 14.65), diffuse cerebral oedema (OR 3.79, 95%CI 1.90–
by ventriculitis (21/110, 19%) and intracranial hemor- 7.97) and hydrocephalus (OR 3.37, 95%CI 1.12–12.95)
rhage (20/110, 18%). At ICU discharge, ventriculitis were associated with unfavorable neurological outcome.
was the most frequent (9/49, 18%), followed by abscess/ Based on the univariate logistic regression, the multi-
empyema (6/49, 12%) and ischemic lesion (5/49, 10%). variate logistic model was built with 16 variables, with a
Illustrations of intracranial complications are described sample of 196 patients. Other indicators of unfavorable
in Additional eFigure 2. outcome identified in our study included chronic alcohol
consumption (aOR 3.10, 95%CI 1.27–7.90), chronic vas-
ICU management and hospital outcomes cular disease (aOR 4.41, 95%CI 1.58–13.63), focal neu-
ICU management and outcomes are described in Addi- rological sign (aOR 2.38, 95%CI 1.11–5.23), presence of
tional eTable 5. A total of 170 patients (72%) received intracranial complication at ICU admission (aOR 2.88,
mechanical ventilation, for a median duration of 7 days 95%CI 1.37–6.21) and CSF leukocyte count < 1000 (aOR
[3–14]. Third generation cephalosporins were the most 4.24, 95%CI 2.11–8.83) (Fig. 2). The distribution of scores
frequently used empirical antibiotic therapy (231/237, on the mRS according to independent predictors of unfa-
97%) and 86% (203/237) received adjunctive steroids. vorable outcome is detailed in Additional eFigure 5.
Reduced susceptibility to β-lactams was low (15%), and Competing risk analysis, with persistent disability (mRS
antibiotic de-escalation was made in 58% cases. Amoxi- score 3–5) as the primary risk and ICU-death as the com-
cillin was the most frequently used antibiotic treatment peting risk, revealed that chronic alcohol consumption
after de-escalation, and the median duration of antibi- was the sole significant variable associated with persis-
otic therapy was 13 days [9–14]. The duration of ICU and tent disability at 90 days (Cause-specific Hazard Ratio
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 6 of 12

Fig. 1 Prevalence of intracranial complications. A. Regardless of the time of hospitalization, B. On ICU admission, C. during ICU stay. *Numbers do not add
up to total and percentage to 100% because one patient may have several complications
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 7 of 12

Fig. 2 Factors associated with poor outcome identified by logistic regression, multivariate analysis

4.26, 95%CI 1.83–9.91), whereas the remaining variables We identified intracranial complications as detected
were associated with mortality (Fig. 3). The Fine and on neuroimaging as a strong indicator of unfavorable
Gray Sub distribution Hazard model adjusted for illness outcome, in the form of death. Most frequent intracra-
severity displayed similar results, with a Sub Distribution nial complications consisted of ischemic lesion and dif-
Hazard Ratio (SDHR) for chronic alcohol consumption of fuse cerebral oedema. The distribution of intracranial
3.58 (95%CI 1.16-11.00). The cause-specific cumulative complications differed depending on the time of diag-
hazards of unfavorable functional outcome with death as nosis. While cerebrovascular complications and diffuse
competing risk according to types of intracranial compli- cerebral edema were highly prevalent at all timepoints,
cation are presented in Fig. 4. ventriculitis was rarely present on ICU admission, and
more prevalent afterwards. In our study, diffuse cere-
Outcomes of excluded patients bral edema was most often detected by CT scan studies,
Post-hoc analyses of the 79 patients staying less than while it is well-known that MRI is a more sensitive tool
48 h in ICU showed that 22 of them were extremely ill for detecting ischemic lesions. This can be explained by
at admission, as reflected by a median SAPS II of 86 the fact that brain CT scans were performed at an early
[7094] and died in ICU. The 57 other patients discharged stage, identifying the most severe patients with increased
alive from ICU within 48 h had a favorable neurological intracranial pressure or in a state of brain death. MRI was
outcome. performed later on in survivors, who did not exhibit the
most severe clinical characteristics. Considering these
Discussion results, we can recommend that CT remains a more
In this multicenter study conducted in adult patients accessible examination allowing for rapid confirmation of
with severe pneumococcal meningitis, we found that a suspicion of increased intracranial pressure with diffuse
intracranial complications, diagnosed on neuroimaging, cerebral edema or a suspicion of brain death. We sug-
were detected in 57% of cases during the entire course gest that for patients with persistent encephalopathy or
of the hospitalization (including intracranial complica- remaining unresponsive after sedation interruption, MRI
tion diagnosed on the first imaging during admission). studies should be proposed to detect other intracranial
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 8 of 12

Fig. 3 Factors associated with poor outcome identified by multivariate competing risk analysis

complications which may have important prognostic once and repeated in case of unexplained neurological
consequences. deterioration during ICU stay.
To the best of our knowledge, this is the first study Our study described a representative population of
specifically investigating the occurrence of intracranial severely affected patients with high compliance to inter-
complications in patients with pneumococcal meningi- national guidelines [17]. We observed the use of empiri-
tis throughout the entire duration of the intensive care cal 3rd generation cephalosporin antibiotic therapy and
process. The diagnosis of intracranial complications may adjunctive steroids in 97% and 86% of cases, respectively.
dramatically impact on the management of the patient. In contrast, a recent retrospective cohort study in the UK
Of note, the identification of infectious complications and Ireland showed that only 57% patients with pneu-
(i.e. brain abscess, ventriculitis), may alter the antibiotic mococcal meningitis received steroids [30]. We did not
regimen and duration of therapy. Other complications observe any association between steroid administration
may sometimes require additional interventions, such as and intracranial complications. Of note, corticosteroid
neurosurgical drainage in case of abscess or empyema, or withdrawal at day 4 was not associated with new intra-
placement of an external ventricular drainage to manage cranial complications, notably vascular events.
hydrocephalus. These interventions may need to be per- Previous studies evaluated the occurrence of intracra-
formed urgently to prevent further neurological damage. nial complications in patients with pneumococcal men-
We found very few patients who underwent neurosurgi- ingitis. A nationwide multicenter study identified a high
cal treatment specifically for treatment of hydrocephalus number of intracranial complications at the acute phase
(n = 3) as the number of this complication was small (14 of the disease, mainly ischemic lesion and diffuse cere-
patients). One of the explications was that hydrocepha- bral edema [9]. Another study conducted in the ICU set-
lus was not significant enough to justify external deriva- ting described lower frequencies of such complications,
tion. The diagnosis of intracranial complications may also as compared to our study [31, 32]. The higher incidence
significantly impact prognostication, in case of severe of intracranial complications observed in our study can
cerebrovascular complications. Considering our results, potentially be attributed to several factors. First, the eval-
a systematic neuroimaging should be considered at least uation of these complications was conducted at various
timepoints, allowing for a comprehensive assessment of
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 9 of 12

Fig. 4 Stacked probability plots. Plots illustrate the state occupation probabilities of being in each state— in ICU, alive and out of ICU without disability,
alive and out of ICU with disability or dead—over the 90 days following ICU admission. A. For any type of intracranial complications, B. For infectious
complications, C. For vascular complications, D. For other intracranial complications

their occurrence. Additionally, a wider range of compli- vascular disease, focal neurological sign, and CSF leu-
cations was included in our analysis, leading to a more kocyte count < 1000 cell/microL at admission, as previ-
extensive categorization. ously described [2, 10]. The association between chronic
Other indicators of unfavorable outcome identified in vascular disease and poor neurological outcome is likely
our study included chronic alcohol consumption, chronic linked to patients’ cardiovascular comorbidities. Using
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 10 of 12

competing risk analyses, we identified chronic alco- Conclusion

hol consumption as the only parameter associated with In severely affected adult patients with pneumococ-
persistent disability at 90 days in survivors, while intra- cal meningitis, intracranial complications diagnosed
cranial complication at ICU admission was associated on neuroimaging at admission or during ICU stay were
with death. Several studies already identified chronic observed in 57% of cases. Intracranial complications
alcohol consumption as a poor outcome variable [2, 33, were independently associated with a higher risk of poor
34]. Chronic alcohol consumption has been linked to functional outcome, in the form of persistent disability
increased host susceptibility to bacterial pneumonia due or death. This study highlights the diagnostic and prog-
to alteration of immune regulation leading to immunode- nostic value of neuroimaging studies performed in this
ficiency [35]. In a nationwide observational cohort study population.
including 696 episodes of bacterial meningitis occurring
in 671 patients, alcoholics were more linked to pneumo- Supplementary Information
coccal meningitis and were at risk for unfavorable out- The online version contains supplementary material available at ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​
g​/​1​0​.​1​1​8​6​/​s​1​3​6​1​3​-​0​2​4​-​0​1​4​0​5​-​z​​​​.​ ​​
come compared to non-alcoholic patients [36].
Our study has several strengths, including a representa- Supplementary Material 1
tive population of severely affected patients. All neuroim-
aging were reinterpreted by an independent neurologist Acknowledgements
who was blinded to the patients’ outcome and radiolo- Antoine Vieillard-Baron, MD, PhD, made substantial contribution to the
gist’s report with an extensive description of intracranial acquisition of data for the work (Department of Intensive Care Medicine,
Ambroise Paré Hospital, Boulogne-Billancourt, France). Eric Maury MD,
complications. We used competing risk analyses to iden- PhD, made substantial contribution to the acquisition of data for the work
tify relevant ICU admission or ICU stay factors associ- (Department of Intensive Care Medicine, Saint Antoine Hospital, Paris,
ated with unfavorable outcome, and we adjusted analyses France. Alexandre Demoule MD, PhD, made substantial contribution to the
acquisition of data for the work (Department of Intensive Care Medicine, Pitié-
for non-neurologic organ failure. Salpêtrière Hospital, Paris, France). Frederic Pene, MD, PhD, made substantial
Our study has some limitations, inherent in its ret- contribution to the acquisition of data for the work (Department of Intensive
rospective design. First, the time of primary endpoint Care Medicine, Cochin Hospital, Paris, France). Michel Wolff, MD, PhD, made
substantial contribution to the critical revision of the manuscript (Department
evaluation was heterogeneous, as only 39% of patients of Intensive Care Medicine, GHU Paris Psychiatrie & Neurosciences, Paris,
discharged alive had an accurate evaluation of functional France).
status on mRS at 90 days (+/-15 days). Of note, most
Author contributions
patients evaluated before 90 days had favorable neuro- Dr Sonneville and Dr Legouy had full access to all the data in the study and
logical outcomes. Second, we excluded patients with ICU take responsibility for the integrity of the data and the accuracy of the data
stays < 48 h, as these patients did not represent a relevant analysis. Concept and design: Legouy, Couffignal, Sonneville. Acquisition,
analysis, or interpretation of data: Legouy, Couffignal, Sonneville. Acquisition
population for neuroprognostication in ICU. Indeed, of data: Legouy, Cornic, Razazi, Contou, Legriel, Garrigues, Buiche, Decavele,
most of short stayers were discharged alive with favorable Benghanem, Rambaud. Drafting of the manuscript: Legouy, Cornic, Couffignal,
neurological outcomes at 90 days, and the remainders Sonneville. Critical revision of the manuscript for important intellectual
content: Legouy, Couffignal, Cornic, Timsit, Sonneville. Statistical analysis:
had a fulminant course and died within 48 h after ICU Couffignal, Cornic, Esposito-Farèse.
admission. Third, neither the timing nor the technique
of neuroimaging was standardized between centers. The Funding
None.
accurate evaluation of intracranial complications on the
day of ICU admission was impossible if there was no Data availability
imaging performed at this time, particularly for ischemic The datasets used and/or analyzed during the current study are available from
the corresponding author on reasonable request.
lesion as the distinction between acute and sequels can
be relatively difficult if only a CT scan was performed.
Thus, immortality bias may have been introduced with Declarations
complications on admission being missed, leading to Ethics approval and consent to participate
classification bias. Nevertheless, neuroimaging is typi- The study protocol named PNEUMATICS was approved for all participating
sites by the French regulatory authorities CNIL (Commission Nationale de
cally initiated in response to neurological decline (e.g., l’Informatique et des Libertés, n°922134) on June 30, 2022 and the local
presence of focal neurological signs, occurrence of epi- ethic committee (CLEA-2021-209). Procedures were followed in accordance
lepsy), which can be inferred as the approximate occur- with the ethical standards of the responsible committee on human
experimentation (institutional or regional) and with the Helsinki Declaration
rence time of intracranial complications. Among patients of 1975.
alive at ICU discharge, underestimation of intracranial
complications may have been less important, as it is more Consent for publication
Not applicable.
likely that physicians do not require brain imaging if they
do not suspect an intracranial complication. Competing interests
RS received research grants from the French Ministry of Health and from LFB.
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 11 of 12

Author details meningitis in adults: should we reconsider anticoagulant therapy? PLoS ONE.
1
Service d’anesthésie-réanimation, GHU Paris Psychiatrie & Neurosciences, 2012;7(9):e45271. https:/​/doi.or​g/10.13​71/j​ournal.pone.0045271.
Paris, France 12. Lucas MJ, Brouwer MC, van de Beek D. Delayed cerebral thrombosis in
2
Centre d’Investigation Clinique, Hôpital Bichat Claude Bernard, AP-HP, bacterial meningitis: a prospective cohort study. Intensive Care Med.
Hôpital Bichat, Inserm CIC 1425, 46, rue Henri Huchard, Paris 2013;39(5):866–71. https:/​/doi.or​g/10.10​07/s​00134-012-2792-9.
75018, France 13. Deliran SS, Brouwer MC, Coutinho JM, van de Beek D. Bacterial meningitis
3
Département d’Epidémiologie, Biostatistique et Recherche, AP-HP, complicated by cerebral venous thrombosis. Eur Stroke J. 2020;5(4):394–401.
Hôpital Bichat, Paris 75018, France https:/​/doi.or​g/10.11​77/2​396987320971112.
4
Médecine intensive réanimation, Hôpital Henri Mondor, Créteil, France 14. Luque-Paz D, Revest M, Eugène F, et al. Ventriculitis: a severe complication of
5
Médecine intensive réanimation, Centre hospitalier d’Argenteuil, Central Nervous System infections. Open Forum Infect Dis. 2021;8(6):ofab216.
Argenteuil, France https:/​/doi.or​g/10.10​93/o​fid/ofab216.
6
Médecine intensive réanimation, Centre hospitalier de Versailles, Le 15. Jim KK, Brouwer MC, van der Ende A, van de Beek D. Subdural empyema in
Chesnay, France bacterial meningitis. Neurology. 2012;79(21):2133–9. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​2​1​2​/​
7
Médecine intensive réanimation, Hôpital Ambroise Paré, Boulogne- W​N​L​.​0​b​0​1​3​e​3​1​8​2​7​5​2​d​0​e​​​​.​ ​​​
Billancourt, France 16. Bodilsen J, Dalager-Pedersen M, van de Beek D, Brouwer MC, Nielsen H.
8
Médecine intensive réanimation, Hôpital de Saint Antoine, Paris, France Risk factors for Brain Abscess: a Nationwide, Population-Based, nested case-
9
Groupe Hospitalier Universitaire APHP-Sorbonne Université, site Pitié- control study. Clin Infect Dis. 2020;71(4):1040–6. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​9​3​/​c​i​d​/​c​
Salpêtrière, Service de Médecine Intensive et Réanimation (Département i​z​8​9​0​​​​.​ ​​​
R3S), Paris F-75013, France 17. van de Beek D, Cabellos C, Dzupova O, et al. ESCMID guideline: diagnosis and
10
Sorbonne Université, INSERM, UMRS1158 Neurophysiologie Respiratoire treatment of acute bacterial meningitis. Clin Microbiol Infect. 2016;22:S37–62.
Expérimentale et Clinique, Paris F-75005, France https:/​/doi.or​g/10.10​16/j​.cmi.2016.01.007.
11
Médecine intensive réanimation, Hôpital Cochin, Paris, France 18. Kasanmoentalib ES, Brouwer MC, van der Ende A, van de Beek D. Hydro-
12
Médecine intensive réanimation, Hôpital Avicenne, Bobigny, France cephalus in adults with community-acquired bacterial meningitis. Neurology.
13
Médecine intensive réanimation, Hôpital Delafontaine, Saint-Denis, 2010;75(10):918–23. https:/​/doi.or​g/10.12​12/W​NL.0b013e3181f11e10.
France 19. Weisfelt M, van de Beek D, Spanjaard L, Reitsma JB, de Gans J. Clinical fea-
14
Université Paris Cité, INSERM U1137, Paris F-75018, France tures, complications, and outcome in adults with pneumococcal meningitis:
15
Médecine intensive réanimation, Hôpital Bichat - Claude Bernard, 6 Rue a prospective case series. Lancet Neurol. 2006;5(2):123–9. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​
Henri Huchard, Paris 75018, France 0​1​6​/​S​1​4​7​4​-​4​4​2​2​(​0​5​)​7​0​2​8​8​-​X​​​​.​ ​​​
16
Université Paris Cité, IAME, INSERM UMR 1137, Paris 75018, France 20. Weller J, Enkirch JS, Lehmann F, Radbruch A, Klockgether T, Zimmermann J.
Early intracranial hemorrhage predicts poor clinical outcome in community-
Received: 24 May 2024 / Accepted: 4 November 2024 acquired bacterial meningitis. Front Neurol. 2022;13:869716. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​
0​.​3​3​8​9​/​f​n​e​u​r​.​2​0​2​2​.​8​6​9​7​1​6​​​​.​ ​​​
21. Hoen B, Varon E, Debroucker T, et al. Management of acute community-
acquired bacterial meningitis (excluding newborns). Short text. Med Mal
Infect. 2019;49(6):367–98. https:/​/doi.or​g/10.10​16/j​.medmal.2019.03.008.
References 22. Meyfroidt G, Kurtz P, Sonneville R. Critical care management of infectious
1. Koelman DLH, Brouwer MC, van de Beek D. Resurgence of pneumococ- meningitis and encephalitis. Intensive Care Med. 2020;46(2):192–201. ​h​t​t​​p​s​:​/​​/​
cal meningitis in Europe and Northern America. Clin Microbiol Infect. d​o​​i​.​​o​r​g​/​1​0​.​1​0​0​7​/​s​0​0​1​3​4​-​0​1​9​-​0​5​9​0​1​-​w​​​​.​ ​​​
2020;26(2):199–204. https:/​/doi.or​g/10.10​16/j​.cmi.2019.04.032. 23. Le Gall JR, Lemeshow S, Saulnier F. A new simplified Acute Physiology score
2. Tubiana S, Varon E, Biron C, et al. Community-acquired bacterial meningi- (SAPS II) based on a European/North American multicenter study. JAMA.
tis in adults: in-hospital prognosis, long-term disability and determinants 1993;270(24):2957–63. https:/​/doi.or​g/10.10​01/j​ama.270.24.2957.
of outcome in a multicentre prospective cohort. Clin Microbiol Infect. 24. Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A
2020;26(9):1192–200. https:/​/doi.or​g/10.10​16/j​.cmi.2019.12.020. practical scale. Lancet. 1974;2(7872):81–4. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​1​6​/​s​0​1​4​0​-​6​7​3​
3. Brouwer MC, van de Beek D. Epidemiology of community-acquired bacterial 6​(​7​4​)​9​1​6​3​9​-​0​​​​.​ ​​​
meningitis. Curr Opin Infect Dis. 2018;31(1):78–84. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​9​7​/​Q​ 25. Rhodes A, Evans LE, Alhazzani W, et al. Surviving Sepsis Campaign: Interna-
C​O​.​0​0​0​0​0​0​0​0​0​0​0​0​0​4​1​7​​​​.​ ​​​ tional guidelines for Management of Sepsis and Septic Shock: 2016. Intensive
4. Lucas MJ, Brouwer MC, van der Ende A, van de Beek D. Outcome in patients Care Med. 2017;43(3):304–77. https:/​/doi.or​g/10.10​07/s​00134-017-4683-6.
with bacterial meningitis presenting with a minimal Glasgow Coma Scale 26. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a
score. Neurol Neuroimmunol Neuroinflamm. 2014;1(1):e9. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​ competing risk. J Am Stat Assoc. 1999;94(446):496–509. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​
1​2​1​2​/​N​X​I​.​0​0​0​0​0​0​0​0​0​0​0​0​0​0​0​9​​​​.​ ​​​ 8​0​/​0​1​6​2​1​4​5​9​.​1​9​9​9​.​1​0​4​7​4​1​4​4​​​​.​ ​​​
5. van de Beek D, de Gans J, Spanjaard L, Weisfelt M, Reitsma JB, Vermeulen M. 27. Beyersmann J, Allignol A, Schumacher M. Competing risks and Multistate
Clinical features and prognostic factors in adults with bacterial meningitis. N models with R. New York: Springer; 2012. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​0​7​/​9​7​8​-​1​-​4​6​1​
Engl J Med. 2004;351(18):1849–59. https:/​/doi.or​g/10.10​56/N​EJMoa040845. 4​-​2​0​3​5​-​4​​​​.​ ​​​
6. van de Beek D, de Gans J, Tunkel AR, Wijdicks EFM. Community-acquired 28. Gray B. (2022). _cmprsk: Subdistribution Analysis of Competing Risks_. R
bacterial meningitis in adults. N Engl J Med. 2006;354(1):44–53. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​ package version 2.2–11, https:/​/CRAN.R​-projec​t.or​g/package=cmprsk
g​/​1​0​.​1​0​5​6​/​N​E​J​M​r​a​0​5​2​1​1​6​​​​.​ ​​​ 29. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP.
7. Sharew A, Bodilsen J, Hansen BR, Nielsen H, Brandt CT. The cause of death in The strengthening the reporting of Observational studies in Epidemiology
bacterial meningitis. BMC Infect Dis. 2020;20(1):182. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​1​8​6​/​s​ (STROBE) statement: guidelines for reporting observational studies. Lancet.
1​2​8​7​9​-​0​2​0​-​4​8​9​9​-​x​​​​.​ ​​​ 2007;370(9596):1453–7. https:/​/doi.or​g/10.10​16/S​0140-6736(07)61602-X.
8. Schut ES, Lucas MJ, Brouwer MC, Vergouwen MDI, van der Ende A, van de 30. Ellis J, Harvey D, Defres S, et al. Clinical management of community-acquired
Beek D. Cerebral infarction in adults with bacterial meningitis. Neurocrit Care. meningitis in adults in the UK and Ireland in 2017: a retrospective cohort
2012;16(3):421–7. https:/​/doi.or​g/10.10​07/s​12028-011-9634-4. study on behalf of the National Infection Trainees Collaborative for Audit and
9. Koelman DLH, Brouwer MC, ter Horst L, Bijlsma MW, van der Ende A, van de Research (NITCAR). BMJ Open. 2022;12(7):e062698. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​1​3​6​/​b​
Beek D. Pneumococcal meningitis in adults: a prospective Nationwide Cohort m​j​o​p​e​n​-​2​0​2​2​-​0​6​2​6​9​8​​​​.​ ​​​
Study over a 20-year period. Clin Infect Dis. 2021;74(4):657–67. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​ 31. Martín-Cerezuela M, Aseginolaza-Lizarazu M, Boronat-García P, et al. Severe
/​1​0​.​1​0​9​3​/​c​i​d​/​c​i​a​b​4​7​7​​​​.​ ​​​ community-acquired Streptococcus pneumoniae bacterial meningitis:
10. Kastenbauer S, Pfister H. Pneumococcal meningitis in adults: Spectrum clinical and prognostic picture from the intensive care unit. Crit Care.
of complications and prognostic factors in a series of 87 cases. Brain. 2023;27(1):72. https:/​/doi.or​g/10.11​86/s​13054-023-04347-3.
2003;126(5):1015–25. https:/​/doi.or​g/10.10​93/b​rain/awg113. 32. Benadji A, Debroucker T, Martin-Blondel G, et al. Cerebrovascular complica-
11. Mook-Kanamori BB, Fritz D, Brouwer MC, van der Ende A, van de Beek D. tions in patients with community-acquired bacterial meningitis: occurrence
Intracerebral hemorrhages in adults with Community Associated bacterial and associated factors in the COMBAT multicenter prospective cohort. BMC
Infect Dis. 2023;23(1):376. https:/​/doi.or​g/10.11​86/s​12879-023-08320-x.
Legouy et al. Annals of Intensive Care (2024) 14:182 Page 12 of 12

33. Bijlsma MW, Brouwer MC, Kasanmoentalib ES, et al. Community-acquired 36. Weisfelt M, de Gans J, van der Ende A, van de Beek D. Community-acquired
bacterial meningitis in adults in the Netherlands, 2006–14: a prospective bacterial meningitis in alcoholic patients. PLoS ONE. 2010;5(2):e9102. ​h​t​t​​p​s​:​/​​/​
cohort study. Lancet Infect Dis. 2016;16(3):339–47. ​h​t​t​​p​s​:​/​​/​d​o​​i​.​​o​r​g​/​1​0​.​1​0​1​6​/​S​1​ d​o​​i​.​​o​r​g​/​1​0​.​1​3​7​1​/​j​o​u​r​n​a​l​.​p​o​n​e​.​0​0​0​9​1​0​2​​​​.​ ​​​
4​7​3​-​3​0​9​9​(​1​5​)​0​0​4​3​0​-​2​​​​.​ ​​​
34. Streharova A, Krcmery V, Kisac P, et al. Predictors of inferior outcome in com-
munity acquired bacterial meningitis. Neuro Endocrinol Lett. 2007;28(Suppl Publisher’s note
3):2–4. Springer Nature remains neutral with regard to jurisdictional claims in
35. Cook RT. Alcohol abuse, alcoholism, and damage to the immune system–a published maps and institutional affiliations.
review. Alcohol Clin Exp Res. 1998;22(9):1927–42.