REVIEWS

INFECTIONS OF THE CNS

Tuberculous meningitis
Robert J. Wilkinson1–3, Ursula Rohlwink4, Usha Kant Misra5, Reinout van Crevel6,
Nguyen Thi Hoang Mai7, Kelly E. Dooley8, Maxine Caws9, Anthony Figaji4, Rada Savic10,
Regan Solomons11 and Guy E. Thwaites7,12 on behalf of the Tuberculous Meningitis
International Research Consortium
Abstract | Tuberculosis remains a global health problem, with an estimated 10.4 million cases and
1.8 million deaths resulting from the disease in 2015. The most lethal and disabling form of
tuberculosis is tuberculous meningitis (TBM), for which more than 100,000 new cases are
estimated to occur per year. In patients who are co-infected with HIV‑1, TBM has a mortality
approaching 50%. Study of TBM pathogenesis is hampered by a lack of experimental models that
recapitulate all the features of the human disease. Diagnosis of TBM is often delayed by the
insensitive and lengthy culture technique required for disease confirmation. Antibiotic regimens
for TBM are based on those used to treat pulmonary tuberculosis, which probably results in
suboptimal drug levels in the cerebrospinal fluid, owing to poor blood–brain barrier penetrance.
The role of adjunctive anti-inflammatory, host-directed therapies — including corticosteroids,
aspirin and thalidomide — has not been extensively explored. To address this deficit, two expert
meetings were held in 2009 and 2015 to share findings and define research priorities. This Review
summarizes historical and current research into TBM and identifies important gaps in our
knowledge. We will discuss advances in the understanding of inflammation in TBM and its potential
modulation; vascular and hypoxia-mediated tissue injury; the role of intensified antibiotic
treatment; and the importance of rapid and accurate diagnostics and supportive care in TBM.

Extrapulmonary
The absolute incidence of tuberculous meningitis (TBM) suggesting that several thousand very young children die
Tuberculosis occurring outside and the overall proportion of meningitis cases that are of TBM every year6. The neonatal Bacillus Calmette–
the lungs. attributable to TBM vary greatly by location, and are Guérin (BCG) vaccine is thought to be 73% effective in
influenced by the overall incidence of tuberculosis, age the prevention of TBM, and its use is estimated to avert
structure, and HIV‑1 seroprevalence within a popula‑ 30,000 childhood cases of the disease annually3, consist‑
tion. Population-based estimates of TBM incidence are ent with the occurrence of around 100,000 cases of TBM
infrequently reported, and are challenging to determine per year overall.
because the diagnosis of TBM is often not microbio‑ In adults, the best-documented risk factor for TBM
logically confirmed. In 2013, a 7‑year national study is HIV‑1 co‑infection. Among HIV-infected individuals
conducted in Germany documented that 422 of 46,349 who live in areas where tuberculosis is highly endemic,
(0.9%) patients with tuberculosis had meningitis, with an the proportion of HIV‑1‑associated meningitis cases
increased risk of TBM in children younger than 5 years attributable to Mycobacterium tuberculosis can exceed
of age (OR 4.90)1. This predisposition towards TBM in 50%7. In addition, in an autopsy study conducted in
young children is commonly reported2,3. In 2015, an esti‑ Kenya, occult, undiagnosed TBM was found in 26%
mated 10.4 million cases of tuberculosis occurred glob‑ of individuals with disseminated HIV-tuberculosis8.
ally, and of 6.1 million incident cases reported, 15% were Individuals with TBM and a HIV‑1 co‑infection have
extrapulmonary4. A large Brazilian study of 57,217 cases of a twofold to threefold increase in relative risk of death
extrapulmonary tuberculosis estimated that meningitis from any cause9,10, with overall mortality around 40%,
7
Oxford University Clinical accounted for 6% of extrapulmonary presentations5. A even in those individuals prescribed antiretroviral ther‑
Research Unit,
764 Vo Van Kiet, Quan 5,
diagnosis of TBM is difficult to ascertain, so the disease apy7,10. Drug-resistant TBM in people co‑infected with
Ho Chi Minh City, Vietnam. might be underreported, but extrapolation suggests that HIV‑1 has a particularly poor prognosis, approaching
gthwaites@oucru.org the global burden of TBM could be at least 100,000 cases 100% mortality11.
doi:10.1038/nrneurol.2017.120 per year. A systematic review of treatment outcomes in This Review capitalizes on knowledge shared in spe‑
Published online 8 Sep 2017 1,636 children with TBM estimated a mortality of 19.3%, cialist meetings held in 2009 and 2015 to summarize

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 1

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Key points Rich and McCordock produced what are now regarded
as the definitive accounts of the extension of tuberculosis
• Tuberculous meningitis (TBM) causes death and disability, with especially high rates of infection to the meninges12. They showed that although
poor outcomes in children and individuals with an HIV‑1 co‑infection TBM might occur as part of disseminated disease, only
• Important risk factors for poor outcome are delayed diagnosis, delayed treatment, a single meningeal or parameningeal granuloma (now
advanced disease, and antitubercular drug resistance termed the Rich focus) was usually found at autopsy and,
• Intracerebral and spinal pathology in TBM is mediated by a dysregulated thus, the point of entry of bacilli into the cerebrospinal
inflammatory response that contributes to meningitis, tuberculoma formation, fluid (CSF) could be located. Rupture of that focus was
arteritis, obstruction of cerebrospinal fluid (CSF) flow, and vascular complications postulated to lead to further extension of inflammation
including stroke throughout the meninges, leading to parenchymal tuber-
• Diagnosis of TBM is insensitive and laborious; clinical scoring algorithms are culoma formation or to vascular pathology characteristic
imperfect and few rigorous evaluations of diagnostics have been performed of the disease. Ischaemia caused by vascular occlusion
• Multidrug antitubercular antibiotic therapy is the mainstay of treatment; however, owing to inflammation both within and around vessels is
CSF penetration is probably a major limitation of these therapies, and evidence compounded by raised intracranial pressure and hydro‑
supporting dosage and treatment combinations is weak cephalus (especially in children, in whom hydrocephalus
• The supportive management of TBM complications, which include hyponatraemia, is nearly invariable13), which can, in turn, compress or
hydrocephalus, hypoxic brain damage and infarction, is poorly understood and cause traction on already inflamed vessels (FIG. 1).
researched, but is vital to outcome
Historically, the most widely used animal model
of TBM consisted of rabbits infected via intracisternal
inoculation of M. tuberculosis, resulting in histologi‑
historical and current research into TBM, and identifies cal, pathological and neurological features consistent
key gaps in our comprehension of the disease. We will with TBM14. However, haematogenous spread cannot
discuss current progress in our understanding of inflam‑ be investigated in this experimental setting and, thus, a
mation in TBM and its potential modulation; vascular detailed understanding of pathogenesis — which should
Miliary and hypoxia-mediated tissue injury; the role of intensi‑ include CNS invasion — cannot be achieved using this
Disseminated micronodular
tuberculosis of the lungs.
fied antibiotic treatment; and the importance of rapid model. Zebrafish have also been used to model TBM:
and accurate diagnostics and supportive care in TBM. Mycobacterium marinum infection of zebrafish via the
Rich focus bloodstream resulted in infection of the CNS and devel‑
The initial intracranial lesion of Pathogenesis opment of associated granulomas in 70% of the fish, with
tuberculous meningitis, as
Macroscopic pathology. Transmission of M. tuberculosis CNS granulomas typically located in close proximity to
described by Arnold Rich.
is airborne and, consequently, the first focus of infec‑ blood vessels15. Infection with a mutant M. marinum
Tuberculoma tion is the lung. Following initial bacterial replication, lacking the homologue of EsxA, a gene associated with
A clinical manifestation of the infection disseminates to the lymph nodes. For M. tuberculosis virulence, resulted in substantial changes
tuberculosis in which tubercles
miliary and extrapulmonary presentations of the disease, in pathology without affecting the ability of the bacte‑
comglomerate into a firm
lump, and so can mimic cancer
dissemination of tuberculosis bacilli must then occur ria to penetrate the blood–brain barrier. High bacterial
tumours of many types in through the blood, either via direct extension of local loads were observed, but no granulomas developed,
medical imaging studies. infection or via lymph nodes. In the 1930s and 1940s, and only small clusters and scattered isolated phago‑
cytes were detected, indicating that EsxA contributes to
pathogenesis in this system.
In the context of disseminated childhood tuberculo‑
Author addresses
sis, the concept of the Rich focus has been re-evaluated16.
1
Department of Medicine, Imperial College London, Norfolk Place, London W2 1PG, UK. Specifically, although TBM can occur when a long-stand‑
2
The Francis Crick Institute, Midland Road, London NW1 2AT, UK. ing Rich focus bursts and discharges its contents into the
3
Wellcome Centre for Infectious Diseases Research in Africa, Institute of Infectious subarachnoid space, TBM commonly manifests in the
Disease and Molecular Medicine and Department of Medicine, University of setting of concurrent disseminated infection in chil‑
Cape Town, Republic of South Africa.
dren. TBM associated with HIV‑1 co‑infection also has
4
Division of Neurosurgery, University of Cape Town, Anzio Road, Observatory 7925,
Republic of South Africa. distinct pathophysiological features. In 1992, a study
5
Department of Neurology, Sanjay Gandhi Postgraduate Institute of Medical Sciences, observed that extrapulmonary tuberculosis — and, in
Rae Bareli Road, Lucknow, Uttar Pradesh 226014, India. particular, TBM — occurred with greater frequency
6
Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 GA Nijmegen, in people with tuberculosis who were co‑infected with
The Netherlands. HIV‑1 (REF. 17). HIV‑1 co‑infection is associated with an
7
Oxford University Clinical Research Unit, 764 Vo Van Kiet, Quan 5, Ho Chi Minh City, increased occurrence of occult meningeal disease8 and
Vietnam. additional extrapulmonary tuberculosis9. A positive
8
Johns Hopkins University School of Medicine, The Johns Hopkins Hospital, M. tuberculosis blood culture was found in up to 40%
1800 Orleans Street, Baltimore, Maryland 21287, USA. patients with newly diagnosed tuberculosis who had
9
Liverpool School of Tropical Medicine, Pembroke Place, Liverpool L3 5QA, UK.
an HIV‑1 co‑infection and severe immunosuppression18.
10
UCSF School of Pharmacy, Department, Bioengineering, 1700 4th Street,
San Francisco, California 94158, UA. HIV‑1 infection could, therefore, predispose individuals
11
Faculty of Health Sciences, Stellenbosch University, Tygerberg Hospital, to more frequent or multiple seeding of the meninges by
Francie van Zijl Drive, Tygerberg 7505, Cape Town, Republic of South Africa. bacilli. Outcome is consistently poor in drug-resistant
12
Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, patients with TBM who have an HIV‑1 co-infection19,20.
University of Oxford, Old Road, Oxford OX3 9FZ, UK. However, the macro­scopic pathology in these individuals

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

seems similar to that found in people without an HIV‑1 to the anti-inflammatory effects of these therapies. In
infection. HIV‑1 co‑infected patients can also experience tissue culture, treatment with dexamethasone mark‑
TBM-associated immune reconstitution inflammatory edly suppresses production of cytokines by microglia
syndrome (TBM-IRIS), which comprises exaggerated and astrocytes30. However, in a clinical trial, although
inflammation in response to M. tuberculosis. In these dexamethasone significantly reduced total protein con‑
circumstances, the introduction of antiretroviral ther‑ centrations in the CSF and marginally reduced CSF
apy after antituberculous chemotherapy is associated concentrations of IFNγ in patients with TBM, other
with the ‘unmasking’ of occult meningeal disease or with cytokines were unaffected31. In a follow‑up study, CSF
worsening of existing TBM. Inflammatory reactions in concentrations of IL‑6 at presentation of TBM were
this context can be intense, although their distribution independently associated with severe TBM, but elevated
and features again do not markedly differ from those CSF levels of inflammatory cytokines were not associ‑
described in people without an HIV‑1 infection19,21. ated with death or disability in HIV‑1-negative patients
with TBM32. Low CSF concentrations of IFNγ were inde‑
Immune and metabolic factors. Bacterial replication pendently associated with death in HIV‑1 co‑infected
must occur in the CNS for TBM pathogenesis to pro‑ patients with TBM, indicating that IFNγ contributes to
ceed. However, the bacillary load in the CSF rarely immunity and survival.
exceeds 100–1000 bacterial colonies per millilitre, and Levels of tumour necrosis factor (TNF) correlate with
viable bacilli are difficult to detect in the majority of the extent of pathology in the rabbit model of TBM, and
individuals. Early studies in experimental animal mod‑ treatment of this model with antibiotics or thalidomide
els showed that the meningitis syndrome and even death (which decreases TNF levels) protects against death33,34.
of tuberculin-sensitized animals could be induced by In children with TBM, CSF levels of TNF and IL‑1β are
meningeal inoculation with dead bacilli22. Much of the elevated35, but no correlation is observed between CSF
tissue damage, therefore, is attributed to a dysregulated levels of TNF and disease stage or response to corti­
host inflammatory response. costeroid therapy36, and another study reported that
Once bacilli have traversed the blood–brain barrier, thalidomide therapy was not associated with decreased
they are taken up by microglia and can also replicate mortality37. Yadav and colleagues reported a positive
in these cells, leading to the induction of microglial correlation between CSF levels of proinflammatory
cytokine and chemokine production23. Conditioned cytokines — including IL‑1β and TNF — and MRI find‑
media from broth cultures of macrophages infected ings, and fractional anisotropy values and post-contrast
with M. tuberculosis induced activation of the NLRP3 signal intensity collected from cerebral cortical regions
(NACHT, LRR and PYD domains-containing protein 3) in patients with TBM38. In a small study, levels of IL‑6,
inflammasome — composed of NLRP3, apoptosis- IL‑10, IL‑1β and IL‑8 were found to be elevated in the
associated speck-like protein and caspase‑1 — in cultured CSF of patients with TBM and declined after 3 months
mouse microglial cells. Furthermore, dexamethasone, a of treatment39. The cytokine levels did not correlate with
corticosteroid used as an adjunctive therapy to reduce stage of meningitis or clinical or radiological outcome,
inflammation in patients with TBM, decreased inflam‑ suggesting that these cytokines cannot be used to infer
masome activation in this culture model through inhibi‑ severity or predict outcome.
tion of reactive oxygen species of mitochondrial origin24. Studies of TBM-IRIS provide further insight into
Elevated levels of reactive oxygen species and malondial‑ inflammatory pathology in HIV-associated TBM.
dehyde have been suggested to mediate tissue damage in Worsening inflammation in TBM-IRIS is associated
childhood meningitis25, similarly suggesting that reactive with high CSF neutrophil counts and M. tuberculosis cul‑
oxygen could contribute to pathology in TBM. ture positivity at TBM presentation40. Multiple cytokines
An 1H-NMR study found an association between are elevated in the CSF compared with the blood, but
decreased CSF levels of glucose and other energy- CSF levels of the neutrophil mediator S100A8/A9 are
related metabolites and elevated CSF levels of lactate in most closely associated with clinical deterioration41.
TBM26. In addition, low levels of glucose and high levels In an unbiased whole-genome analysis of peripheral
of lactate in the CSF were associated with death in an blood transcriptomic response, patients with TBM-IRIS
independent study of prognosis in adults with TBM in exhibited a substantially increased number of transcripts
Vietnam27. A 2016 study found a relationship between associated with canonical and noncanonical inflammas‑
hyponatraemia and the presence of basal exudates , omes after early antiretroviral treatment, compared with
implicating pituitary involvement in hyponatraemia28. non-IRIS controls. This difference suggests a dominant
Indeed, in a systematic investigation of pituitary dys‑ role for the innate immune system in the pathogenesis
function in TBM that included MRI, 13.3% of 75 cases of TBM-IRIS42. In addition, paradoxical worsening of
Basal exudates
had evidence of pituitary abnormality, 42.7% had rela‑ TBM after treatment occurs in patients without an HIV
An inflammatory reaction to
tuberculosis in the basal tive or absolute cortisol insufficiency, 30.7% had central infection who have drug-sensitive TBM. In a study of
cisterns of the brain. hypothyroidism, and 49.3% had hyperprolactinaemia29. 34 patients with definite TBM at the time of first treat‑
The occurrence of cortisol deficiency in TBM is ment who were followed‑up with cranial MRI at 3 and
Paradoxical worsening noteworthy given that synthetic corticosteroids are the 6 months after treatment initiation, paradoxical worsen‑
The worsening of a tuberculosis
lesion during otherwise
best-established adjunctive therapy for this condition. ing was observed in 22 individuals (64.7%), more than
effective antirtubercular or The beneficial effect of these agents might be partially half of whom remained clinically asymptomatic, which
antiretroviral therapy. metabolic, although most attention has been directed suggests that clinically silent radiographic deterioration

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 3

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a Blood
e
Bacteria
Neutrophil
Lymphocyte
Monocyte
Tight
junctions

Basement Endothelial
membrane cells

Astrocyte
Brain Microglial cell

b Uninfected
cells

Diapedesis
of infected
cells
f
Cytokines
Infected
microglial cell

Disrupted
endothelium
and basement
membrane
Diapedesis
of uninfected
cells
Development of
a rich focus with g
a cuff on innate
and specific
B cells and T cells

Cytokines

d Arachnoid mater

White
matter

Grey
matter
CSF
Pia mater
Rupture focus with
spread of infection to
meninges and CSF

Nature Reviews | Neurology

is common43. Furthermore, in a study conducted in pathogenesis of TBM. Nevertheless, an acquired type 1
Vietnam, tuberculomas developed in 74% of patients T-helper cell response to M. tuberculosis antigens, char‑
with TBM during treatment; these tuberculomas were acterized by IFNγ secretion, occurs in the CSF during
associated with long-term fever, but not with relapse or TBM, as documented by studies aimed primarily at
poor clinical outcome44. finding diagnostic markers for TBM32,45. Most such diag‑
These studies suggest that the induction of inflam‑ nostic assays are optimized to detect terminally differ‑
masomes and activation of an innate cytokine response entiated CD4+ T cells, despite the fact that CD8+ T cells
leading to influx of phagocytes are key processes in the with T-effector memory and terminally differentiated

4 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

◀ Figure 1 | Pathogenesis of tuberculous meningitis. a | Bacilli could reach brain blood Immunohistochemical staining of brain tissue from
capillaries within cells or as extracellular bacilli; the precise mechanism is unknown. patients who had TBM demonstrated the presence of
b | The endothelium itself can be infected, or infected cells can adhere and undergo VEGF in the inflammatory mononuclear cells of the
diapedesis. Both processes result in breakdown of tight endothelial junctions and the dense fibroconnective tissue in the subarachnoid space
basement membrane. c | Microglial cells can become infected, and these cells, together
and surrounding vasculitic lesions. In addition, CSF
with infiltrating cells, produce inflammatory chemoattractants that result in further
breakdown of the blood–brain barrier and influx of uninfected cells, including innate and
levels of VEGF were found to be associated with the
specific T and B lymphocytes. d | The nascent granuloma might rupture via necrosis, presence of TBM and MRI evidence of infarction, but
leading to meningeal and intracerebral dissemination of infection. e | Confocal not with death54. Furthermore, CSF levels of hepatocyte
microscopy of a lymph node from an individual infected with HIV‑1, showing localization growth factor (HGF) are higher in patients with TBM
of Mycobacterium tuberculosis (green) in or near endothelial cells and in infected cells than in those with bacterial meningitis, suggesting that
(nucleus blue) within a blood vessel (lining red). f | Consequences of meningeal other growth factors contribute to pathogenesis55.
dissemination of infection. Endoscopic third ventriculostomy in a child illustrates the
cistern of the subarachnoid space beneath the third ventricle, showing no visible normal Bacillary and host genetic factors. M. tuberculosis can
anatomy, with the basilar artery, perforators and brainstem covered by tuberculous be divided into seven lineages classified as ‘ancient’ (line‑
exudate. g | T1 contrast MRI, showing multiple parenchymal granulomas in the basal
ages 1, 5, 6 and 7) or ‘modern’ (lineages 2, 3 and 4)56. One
ganglia and cortex — one of the consequences of intracerebral dissemination. CSF,
cerebrospinal fluid. Part e republished with permission of American Society for Clinical
hypothesis is that some lineages might more frequently
Investigation, from Lerner, T. R. et al. Lymphatic endothelial cells are a replicative niche cause or affect the severity of meningeal infection. In
for Mycobacterium tuberculosis J. Clin. Invest. 126, 1093–1108 (2016). Permission children with TBM in China, lineage 2 was twice as
conveyed through Copyright Clearance Center, Inc. common as other prevalent lineages2, and in Vietnam,
the Thr597Cys polymorphism in the Toll-like receptor 2
(TLR2) gene was associated with a slightly elevated risk
phenotypes46, as well as mycobacterial antigen-reactive of tuberculosis caused by lineage 2 strains57. In a sub‑
Vγ9–Vδ2 T cells that secrete interferon and TNF, are group of 122 patients in Vietnam who were co‑infected
also found in the CSF of patients with TBM47. A relation‑ with HIV‑1 and TBM, patients infected with lineage 2
ship between high levels of IFNγ and the formation of strains had lower mortality than those infected with
intracerebral tuberculomas has been suggested48. C‑X‑C lineage 1 strains11. However, in Thailand, no differences
motif chemokine 10 (also known as 10 kDa IFNγ- in mortality were found between lineages 1, 3 and 4
induced protein, or IP‑10) and C‑X‑C motif chemok‑ among 184 patients with CSF isolates of M. tuberculosis58.
ine 9 (also known as monokine induced by IFNγ, or In children living in Cape Town, South Africa, lineages 2
MIG) are compartmentalized at the site of disease and and 4 showed comparable propensity to cause extrapul‑
decrease after treatment49. However, substantial immune monary disease in general, and TBM in particular59. In
activation and blood–brain barrier abnormalities are still children with tuberculosis in China, a single nucleotide
apparent after 60 days of treatment27. polymorphism in the PE_PGRS33 gene of M. tuberculo-
Although cytokines can mediate inflammatory reac‑ sis was associated with an increased risk of developing
tions, these molecules do not directly cause tissue dam‑ TBM, suggesting that even minor bacillary variation can
age. Proteolytic matrix metalloproteinases (MMP) have influence pathogenesis60.
attracted interest as potential mediators of this damage. Many immune response genes are under polymor‑
The kinetics of CSF MMPs and their endogenous tissue phic genetic influence. Several studies conducted over
inhibitors were studied in 37 HIV-uninfected adults with the past 10 years indicate that polymorphisms in host
TBM in Vietnam who were recruited to a randomized response genes (such as TLR9, SIGIRR, SPN, TIRAP,
controlled trial of adjuvant dexamethasone50. This ther‑ TLR2 and MRC1) can influence the immune response
apy resulted in a substantially reduced CSF concentra‑ and, thus, predispose an individual to TBM (TABLE 1).
tion of MMP‑9 after 5 days of follow‑up. The MMP‑9 Consistent with data from cytokine studies, many of the
concentration correlated with the CSF neutrophil count. reports document that polymorphism in innate immune
Paradoxically, suppression of microglial secretion of response receptors or signalling pathways influences the
72 kDa type IV collagenase (also known as MMP‑2) by susceptibility of an individual to TBM itself, or the form
signalling pathways activated by tuberculosis-infected in which it manifests. A striking example is a single
monocytes involves the proinflammatory mediators nucleotide polymorphism in the leukotriene A4 hydro‑
TNF, MAP kinase 14 and nuclear factor‑κB, in addition lase (LTA4H) promoter, which alters the transcriptional
to a novel caspase-8‑dependent pathway51. activity of this gene and, thus, influences the balance
Evidence also indicates a contribution from growth between proinflammatory leukotriene B4 and immuno­
factors — particularly those involved in angiogenesis suppressive lipoxin A4. In TBM, this polymorphism is
— in the vascular pathology observed in TBM. A associated with inflammatory cell recruitment, patient
study of serum and CSF from 56 children with and 55 survival and response to adjunctive anti-inflammatory
children without TBM found that increased levels of therapy61.
vascular endothelial growth factor (VEGF) were char‑ These intriguing findings, which introduce the pos‑
acteristic of TBM52. In a study examining patients with sibility of personalised anti-inflammatory TBM treat‑
meningitis in Japan, serum and CSF levels of VEGF ment, were replicated in a cohort of adults with TBM in
were substantially higher in TBM than in other forms Vietnam62. All participants received corticosteroids in
of meningitis. Decreases in CSF levels of VEGF paral­ accordance with current clinical guidelines. The LTA4H
leled clinical improvement in patients with TBM53. genotype influenced the pretreatment intracerebral

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 5

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Table 1 | Genetic polymorphisms associated with TBM

Gene Function Location Polymorphisms Association Refs
SPN Surface glycoprotein (CD43) Vietnam rs17842268 rs17842268 associated with decreased 171
involved in Mycobacterium rs12596308 survival from TBM. Both polymorphisms
tuberculosis adhesion and associated with focal neurological deficit at
proinflammatory cytokine TBM presentation
induction
MBL2 Lectin involved in recognition South Africa Gly54Asp allele Protection from TBM 172
and uptake of M. tuberculosis
TLR2 Innate pattern recognition India Arg753Gln and No association 173
receptor Pro631His
China rs3804099 (597T>C) Associated with pulmonary tuberculosis but 174
not meningitis
Vietnam rs3804099 (597T>C) Associated with TBM caused by lineage 2 57,175
strains
TLR9 Innate pattern recognition Vietnam rs352142 TBM rather than PTB 176
receptor
TIRAP Toll adaptor protein Vietnam and 558C>T allele Associated with TBM 177,178
South Africa
PKP3–SIGIRR–ANO9 Negative regulator of Toll-like Vietnam rs10902158 TBM and PTB 179
gene region receptor/IL‑1R signalling rs7105848
rs7111432
VDR Vitamin D receptor India ApaI and TaqI Heterozygous (TC) and mutant (CC) 173
restriction sites genotypes of Taq1 VDR SNP significantly
associated with TBM
LTA4H Leucotriene A4 hydrolase Vietnam rs17525495 High (TT) and low (CC) inflammatory 61
homozygous forms associated with death
from TBM; dexamethasone protects from
death in TT but not CC genotype
Indonesia rs17525495 No significant association with CSF 63
leukocytes or patient survival
CCL2 Chemokine China rs4586 C/T TT genotype associated with lower CSF 180
mononuclear leukocyte counts in TBM
PTB, pulmonary tuberculosis; SNP, single nucleotide polymorphism; TBM, tuberculous meningitis.

inflammatory phenotype and was independently associ‑ paraplegia) in around 10% of patients. Death is nearly
ated with survival in 446 individuals with TBM who did certain unless antitubercular treatment is provided.
not have an HIV infection. By contrast, LTA4H geno­ The clinical features of TBM are influenced by the
type did not influence the inflammatory phenotype or immune response against M. tuberculosis. Very young
survival in 326 adults with TBM and an HIV co‑infec‑ children (<1 year of age) and those with advanced HIV‑1
tion, with the possible exception of those with peripheral co‑infection are highly susceptible to M. tuberculosis,
blood CD4+ cell counts >150 × 106/μl. which frequently leads to uncontrolled extrapulmonary
In an Indonesian study, LTA4H genotype was not asso‑ dissemination and meningitis. In these groups, the pres‑
ciated with survival in 427 HIV-negative patients with entation of TBM can be abrupt, and can rapidly progress
TBM, all of whom received corticosteroids, and did not to severe coma and prostration with high mortality;
influence patient survival in sensitivity analyses63. Thus, furthermore, these individuals have an increased risk of
uncertainty remains as to whether the LTA4H geno­type active tuberculosis in other organs9,65.
determines TBM inflammatory pathophysiology in all Age and HIV‑1 co‑infection also influence the lab‑
populations. oratory features of TBM66. The CSF of individuals with
TBM typically has 150–1000 leukocytes per μl, with a
Clinical features mixed population of neutrophils and lymphocytes
Effects of age and HIV‑1 co‑infection. TBM has a non‑ (although lymphocytes usually predominate), elevated
specific prodrome that gradually evolves to include protein (0.8–2.0 g/dl), and CSF:plasma glucose ratios of
more-recognizable symptoms and signs of meningitis, <0.5 in 90% of individuals. HIV‑1 co‑infection can alter
such as headache, fever, vomiting and neck stiffness64. the CSF inflammatory profile, with either no leukocytes
Without treatment, meningitic symptoms become pro‑ present, or large numbers (>1000 cells per μl) of neutro‑
gressively more dominant, consciousness falls, and focal phils mimicking acute pyogenic bacterial meningitis67.
neurological deficits occur, with cranial nerve palsies — The outcomes of TBM are better the earlier treat‑
largely in the fifth and third cranial nerves — in around ment is started. The Medical Research Council (MRC)
50% of patients and limb weakness (either hemiplegia or disease severity grade remains the most widely used

6 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Box 1 | The Medical Research Council tuberculous meningitis severity grade resonance angiography result at disease presentation
strongly predicts an infarct-free disease course78.
In 1948, Medical Research Council investigators in the UK graded tuberculous Spinal cord disease is common in people with TBM
meningitis for the first trial of streptomycin as: and often goes undetected: one imaging study of chil‑
• Early — no clinical signs of meningitis or focal neurology and fully conscious dren with TBM and hydrocephalus found that 76% had
• Medium — patient’s condition falls between early and advanced spinal disease79. Definition of the full extent of the CNS
• Advanced — extremely ill, in a deep coma disease is important for treatment and prognosis in peo‑
With the introduction of the Glasgow Coma Scale (GCS) in 1974, this classification was ple with TBM, and the detection of extraneural disease
modified to: — for example, in the lymph nodes — could offer alter‑
• Grade I (GCS score 15; no focal neurological signs) native diagnostic sampling opportunities in those with
• Grade II (GCS score 11–14, or 15 with focal neurological signs) diagnostic uncertainty.
• Grade III (GCS score ≤10)
Diagnosis
Microbiological diagnosis. The use of microscopy and
method to define TBM disease severity68 (BOX 1), and Ziehl–Neelsen staining to detect of acid-fast bacilli in fluids
strongly predicts outcome6,69,70. The rate of death or or tissues from affected organs has long been the main‑
severe neurological disability in MRC grades I, II, and stay for the rapid diagnosis of tuberculosis. However, the
III are approximately 15%, 30% and 50%, respectively, detection of <10,000 bacteria per ml remains consist‑
in HIV-1-negative individuals, rising to 25%, 50%, and ently difficult owing to the low numbers of bacilli pres‑
80%, respectively, in people co‑infected with HIV‑1. The ent and time required to find them by microscopy. This
widespread availability of antiretroviral treatment has impediment renders microscopy around 50% sensitive
improved outcomes in HIV‑1 co‑infected individuals for tuberculosis associated with high bacterial burdens
with TBM; however, outcomes remain poor. (such as cavitory pulmonary tuberculosis in adults)
Disease caused by multidrug-resistant M. tuberculosis and 10–20% sensitive in paucibacillary disease (such as
— that is, resistant to at least rifampicin and isoniazid meningitis).
— confers a major risk of death11,71. Before the introduc‑ The sensitivity of CSF microscopy can be increased
tion of rapid molecular tests that can identify rifampicin from 10–20% to >50% by simple measures80. A large vol‑
resistance at TBM disease presentation, almost all ume of CSF (~10 ml), centrifuged at 3,000 g and exam‑
patients died before the results of a conventional drug ined for 30 min by an experienced microscopist, can lead
susceptibility test were returned and second-line treat‑ to the detection of bacteria in >80% of TBM cases81, and
ment could be initiated. Other important risk factors for also increases the yield of culture. However, these meth‑
death from TBM include extremes of age, and interrup‑ ods are rarely practical where they are most needed in
tions to first-line antitubercular treatment (for example, resource-poor settings, and rapid diagnostic methods for
because of drug-induced adverse events)72. TBM with improved accuracy are required.
Culture of M. tuberculosis from patient specimens is
Radiographic appearance. Basal meningeal exudates, more sensitive than microscopy for diagnosis, but takes
infarcts, tuberculomas and hydrocephalus are all com‑ at least 10 days in liquid media and up to 8 weeks on
mon features of TBM73. These features can occur in solid media, and ideally requires a biosafety level 3 lab‑
isolation or in combination, and might not be detected oratory. Clinical decision-making, especially for TBM,
radiographically until the disease becomes advanced. cannot wait this long, and few centres have the required
CT and MRI are widely used for detection of these facilities. Innovative methods for the early detection
signs, although the latter has superior resolution, espe‑ of M. tuberculosis in liquid culture media have been
cially for the detection and definition of infarcts and developed, but only the microscopic observation drug
tuberculomas. susceptibility (MODS) assay has been investigated for
New imaging modalities can assist in TBM diagnosis the diagnosis of TBM82. This method uses microscopy
and management. Diffusion-weighted MRI can detect to detect the early M. tuberculosis-specific ‘cords’ of
the early changes of cerebral ischaemia and borderzone bacterial growth and was found to provide results in a
necrosis, and is more sensitive than conventional MRI median of 6 days when used on CSF, with a sensitivity
for the detection and localization of ischaemia and of around 65%.
infarcts74,75.
The importance of infarction to long-term outcomes Molecular assays. The insensitivity, slow speed and
has led to much interest in cranial vessel imaging. CT expense of conventional bacteriology have provided the
angiography has been used to define lesions in the ante‑ impetus to develop tests based on nucleic acid ampli‑
Ziehl–Neelsen staining
rior and posterior cerebral circulation, and has demon‑ fication to detect M. tuberculosis-specific molecules. A
A technique to visualize
Mycobacterium tuberculosis strated that the supraclinoid portion of the internal plethora of in‑house nucleic acid amplification tests have
directly by microscopy in carotid artery and proximal portions of the anterior cer‑ been developed and tested, but few have been subject to
pathological samples. ebral and middle cerebral arteries are most commonly repeated or well-designed validation83. GeneXpert MTB/
affected in TBM76. Magnetic resonance angiography RIF (manufactured by Cepheid) is a commercial, real-
Paucibacillary
Disease associated with very
has demonstrated vascular lesions, which are related time PCR-based assay for the detection of M. tuberculosis
low numbers of bacteria in to hydrocephalus, infarction and poor outcome, in in clinical specimens, and also detects mutations associ‑
clinical specimens. around 50% of patients with TBM77. A normal magnetic ated with rifampicin resistance. This assay was endorsed

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 7

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

by the WHO in 2010, and is being adopted globally as a accuracy. Specificity is enhanced when the assays are
front-line diagnostic test. Three studies have investigated used on CSF, but large volumes are required (>2 ml) and
the role of GeneXpert in the diagnosis of TBM84–86, and indeterminate results are common (up to 15%).
found the assay to be around 60% sensitive and nearly Two systematic reviews and meta-analyses have
100% specific. Testing of centrifuged CSF deposit from examined the diagnostic accuracy of CSF levels of
a large volume of CSF enhances sensitivity. The WHO adeno­s ine deaminase in TBM 94,95. For adenosine
has recommended GeneXpert as an essential diagnostic de­aminase values from 1–4 U/l, the sensitivity and spec‑
test in people with suspected TBM, but the results must ificity were >93% and <80%, respectively, compared
be interpreted with caution87,88, and it should be used as with >96% and <59% for values >8 U/l. None of the
a ‘rule‑in’ rather than a ‘rule-out’ test. Further investiga‑ cut-off values could reliably discriminate between TBM
tions searching for additional evidence of TBM should and bacterial meningitis. The different methods used to
always be sought, and antitubercular treatment should measure adeno­sine deaminase and the heterogeneity of
be started without microbiological proof in individuals data limit standardization of this test in routine practice.
clinically suspected of having TBM. A second generation
GeneXpert test (Ultra) is available that relies on detec‑ Diagnostic imaging. Brain imaging has long been part of
tion and amplification of a multicopy gene target, and a the diagnostic assessment of TBM. Numerous case series
2017 WHO report suggested 95% sensitivity for TBM89. have described the common radiographic appearances
The limitations of nucleic acid amplification have of basal meningeal exudates, hydrocephalus, infarcts and
led to investigation of the M. tuberculosis glycolipid tuberculomas (FIG. 2). However, few studies have defined
lipoarabinomannan as a diagnostic marker in CSF. A the diagnostic performance of these features.
comparison of lipoarabinomannan detection by lateral Early studies suggested that basal meningeal exu‑
flow assay (LFA) or ELISA with GeneXpert in CSF from dates, identified by contrast-enhanced CT imaging,
an autopsy cohort of 91 adults with confirmed TBM and were specific for TBM and predicted poor outcome96.
an HIV co‑infection found the sensitivity of lipoara‑ Subsequent investigators reported that post-contrast
binomannan LFA to be 75%, compared with 43% for basal exudates detected by CT imaging were both
ELISA and 100% for Gene Xpert MTB/RIF, suggesting sensitive (~90%) and specific (~95%) for the diagno‑
that lipoarabinomannan LFA could be a useful test90. sis of TBM in children97. Pre-contrast hyperdensity
However, others have found lipoarabinomannan LFA in the basal cisterns might be an even more accurate
to be substantially less sensitive in practice than was predictor of TBM97, but further research is needed
suggested by this study, especially in those without an to validate this finding. Various objective criteria for
HIV‑1 infection91. In HIV co‑infected individuals, per‑ post-contrast CT basal enhancement have been pro‑
formance could be enhanced by combining the test with posed, most of which are specific but lack sensitivity98.
a clinical diagnostic prediction rule92. A study evaluating these criteria in adults again
found high specificity (90–100%) but low sensitivity
Diagnosis based on the host response. The difficulty (0–30%)99. Furthermore, inter-reader variability was
of detecting M. tuberculosis in the CSF has driven high, limiting clinical utility.
interest in whether specific immune responses can Infarcts and hydrocephalus revealed by CT brain
aid TBM diagnosis. IFNγ release assays depend on the imaging lack diagnostic specificity for TBM, as other
ability of T lymphocytes from M. tuberculosis-infected infectious and noninfectious diseases can cause simi‑
individuals to produce IFNγ when stimulated with lar features. An increased specificity (95–100%) can be
M. tuberculosis-specific antigens. A 2016 systematic achieved via detection of the combination of hydro‑
review and meta-analysis found that the overall sensi‑ cephalus, basal enhancement and infarction; however,
tivities of blood and CSF IFNγ release assays were 78% these features can be absent in early disease, which
and 77% respectively, with 61% and 88% specificity93. reduces diagnostic sensitivity to around 40%. MRI can
These data suggest that these assays have only moderate provide additional diagnostic information100; for exam‑
ple, gadolinium-enhanced MRI is superior to CT for the
visualization of small leptomeningeal tuberculomas101,
a b c and diffusion-weighted MRI improves the detection
of small areas of ischaemia or early infarction74. In one
study102, MRI revealed 55 cases of stroke — 54 ischaemic
and one haemorrhagic — in 122 patients with TBM.
Infarcts were present in the basal ganglia (n = 30), the
thalamus (n = 9), the brainstem (n = 10), the cerebral
cortex (n = 27) and the cerebellum (n = 4). 29 patients
experienced multiple strokes. 38 had infarcts in the
anterior circulation, seven had infarcts in the poste‑
Figure 2 | MRI scans from a patient with stage II tuberculous meningitis. The patient
rior circulation and the remaining ten had infarcts in
Nature Reviews | Neurology both areas102. However, the diagnostic sensitivity and
had presented with fever for 15 days and altered sensorium for 1 day. a | Granuloma in T1
contrast axial section. b | Anterior cerebral artery territory infarction in diffusion-weighted specificity of these features for TBM are undefined.
sequence. c | Narrowing right anterior cerebral artery and occlusion of left anterior The main limitation of both imaging modalities is
communicating artery. that around 30% of individuals at an early stage in the

8 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

TBM disease course (MRC grade I) have normal brain case definition was not designed to be used as a clinical
CT scans, and around 15% have normal brain MRI scans. diagnostic tool, and was shown to perform poorly when
Furthermore, age and HIV‑1 co‑infection status influ‑ used in this manner115.
ence the appearance of the brain on imaging. Children
with TBM are more likely to exhibit hydrocephalus Management
than are adults with TBM, and basal enhancement is Antimicrobial therapy. Antitubercular treatment before
often less prominent in people co-infected with HIV‑1, the onset of coma is the strongest predictor of survival
especially those with very low CD4+ cell counts103. from TBM. However, the best combinations, doses, fre‑
Follow‑up imaging during treatment can pro‑ quencies and durations of these drugs have not been
vide additional useful diagnostic information, as well determined, despite their importance to a successful out‑
as important evidence concerning complications104. come. The treatment of drug-susceptible TBM is based
Imaging of organs other than the brain can also yield on the regimens used against pulmonary tuberculosis.
important diagnostic information. Concomitant tuber‑ The WHO recommends the same regimen of 2 months
culosis of the spinal cord can occur in more than half of of rifampicin, isoniazid, pyrazinamide and ethambutol
individuals with TBM, and is commonly overlooked79 followed by 10 months of rifampicin and isoniazid, for
unless MRI or PET–CT scans are performed105. Plain all patients. The recommended doses and frequencies
chest radiography is normal in around 50% of people for TBM treatment are the same as for pulmonary tuber‑
with TBM, and the remaining individuals have a spec‑ culosis, despite concerns that rifampicin — in particu‑
trum of changes suggestive of previous or currently lar — does not penetrate the blood–brain barrier well,
active tuberculosis 106. Only the chest radiographic and concentrations of this agent in the CSF are 10–20%
appearance of miliary tuberculosis indicates dissemi‑ of those achieved in plasma116,117 (TABLE 3). In addition,
nated tuberculosis and the strong likelihood of TBM. ethambutol does not cross the blood–brain barrier, even
Radiographic evidence of tuberculosis at other sites, when it is inflamed. Children generally require higher
acquired by conventional or PET–CT of the chest, doses per kg of body weight than do adults to achieve
abdomen and pelvis, can provide supportive diagnostic similar exposures to antituberculosis drugs, and this dif‑
information and offers low-risk opportunities to biopsy ference is reflected in the current paediatric tuberculo‑
the affected tissues. sis treatment guidelines118,119. Paediatric TBM treatment
guidelines mirror those of adults in terms of regimen
Clinical scoring systems. Several studies have described composition and treatment duration. In practice,
the clinical features that are most predictive of TBM, paediatric treatment regimens vary, as no clinical trials
leading to the development of diagnostic scoring sys‑ data are available to guide drug combination and dose
tems (TABLE 2). The findings vary according to the selection in paediatric TBM.
population in which they were derived, with age and Whether rifampicin — the key sterilizing drug
HIV status accounting for much of this variation. in tuberculosis treatment — should be used at doses
Nevertheless, these scoring systems serve to highlight >10 mg/kg for the treatment of TBM is controversial.
the key distinguishing features of TBM: long-term Paediatricians in Cape Town have long used a hyperin‑
symptoms (>5 days), lower numbers of CSF leukocytes tensive regimen, consisting of 20 mg/kg rifampicin, iso‑
(<1000 cells per mm3) than are typically observed in niazid and ethionamide, with 40 mg/kg pyrazinamide, all
other forms of bacterial mening itis, CSF leukocytes given for 6 months. They report few adverse events and
comprised of <90% neutrophils, elevated CSF protein excellent outcomes120. Pharmacokinetic modelling studies
(>100 mg/dl), and a low CSF:blood glucose ratio (<0.5) have suggested that in childhood TBM, rifampicin doses
compared with uninfected individuals. of at least 30 mg/kg orally, or 15 mg/kg intravenously,
A major limitation of these clinical diagnostic rules could be required to achieve sufficient CSF drug expo‑
is that their performance is variable in different popu‑ sure and bacterial killing to reduce mortality121. Whether
lations and settings, and few of these scoring systems the same applies to adults is currently uncertain. A trial
have been externally validated. A diagnostic rule devel‑ conducted in 60 adults with TBM in Indonesia found
oped in Vietnam is currently the only one to be been that mortality was nearly halved if rifampicin was given
tested in different populations. This rule was originally at a dose of 600 mg intravenously, compared with the
described to be 86% sensitive and 79% specific for standard dose of 450 mg orally, resulting from a three‑
TBM diagnosis in Vietnamese adults107, and subsequent fold increase in systemic rifampicin exposure for the first
studies in Turkey108, Vietnam109, India110, China111 and 2 weeks of combination antituberculosis treatment122.
Colombia112 reported sensitivities >90% and specificities A strong concentration–effect relationship was found,
ranging from 50–90%. However, a serious limitation of with higher rifampicin exposure being associated with
the Vietnam rule was exposed by a study in Malawi of 86 better TBM survival. Minimum rifampicin target values
HIV‑1 co‑infected adults with meningitis113. In this pop‑ were derived from exposure–response curves, and only
ulation, the rule was only 78% sensitive and 43% specific, 38% of patients receiving 600 mg intravenously reached
with cryptococcal meningitis accounting for most of the these target rifampicin exposures, suggesting that even
false-positive results. higher doses are required to optimize rifampicin dosing.
A consensus case definition for TBM has been pub‑ However, a subsequent a trial in 817 adults with TBM
lished to help standardize clinical TBM research and in Vietnam failed to show any survival benefit of using
enable direct comparison of studies114. However, the increased doses of rifampicin (15 mg/kg given orally

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 9

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Table 2 | Published clinical diagnostic rules for the diagnosis of TBM

Publication Population from which Case comparison Predictors of TBM
rule derived
Kumar et al.181 Children from India TBM versus other meningitis • Symptoms ≥7 days
• Optic atrophy
• Focal neurological deficit
• Extrapyramidal movements
• CSF leukocytes <50% neutrophils
Thwaites et al.107 Adults from Vietnam TBM versus bacterial meningitis • Age <36 years
• Blood leukocytes <15 × 109/l
• Symptoms ≥6 days
• CSF leukocytes <750/mm3
• CSF neutrophils <90%
Youssef et al.182 Children and adults from TBM versus bacterial meningitis • Symptoms >5 days
Egypt • CSF leukocytes <1000/mm3
• Clear CSF
• CSF lymphocytes >30%
• CSF protein >100 mg/l
Cohen et al.183 Adults from Malawi (77% TBM versus cryptococcal • Low CSF opening pressure
HIV-infected) meningitis • Neck stiffness
• Raised CSF leukocytes
• Low Glasgow Coma Scale score
• High fever
Patel et al.92 Adults from South Africa TBM versus other meningitis • CSF:blood glucose ratio ≤0.2
(84% HIV-infected) • CSF lymphocytes >200/mm3
• CD4+ cell count <200 × 106/l
• Negative cryptococcal antigen test
Hristea et al.184 Adults from Turkey TBM versus viral meningitis • Symptoms ≥5 days
• MRC grade II or III
• CSF:blood glucose ratio <0.5
• CSF protein >100 mg/dl
Vibha et al.110 Adults from India TBM versus bacterial meningitis • Living in a rural area
• Symptoms ≥6 days
• Cranial nerve palsy
• Hemiplegia
• Clear CSF
• CSF neutrophils <75%
Dendane et al.185 Adults from Morocco TBM versus bacterial meningitis • Female sex
• Symptoms ≥10 days
• Focal neurological deficits
• Blood leukocytes <15 × 109/l
• Plasma sodium <130 mmol/l
• CSF leukocytes <400/mm3
Zhang et al.186 Adults from China (all TBM versus cryptococcal • Female sex
HIV-uninfected) meningitis • Reduced consciousness
• No visual or hearing loss
• Evidence of extraneural
tuberculosis
• CSF leukocytes ≥68/mm3
• CSF protein >0.91 mg/dl
Qamar et al.187 Children from Pakistan TBM versus bacterial meningitis • Hydrocephalus on brain CT
• CSF leukocytes <800/mm3
• CSF protein:glucose ratio ≥2
CSF, cerebrospinal fluid; MRC, Medical Research Council; TBM, tuberculous meningitis.

versus 10 mg/kg given orally in the control arm) with the The problem of poor blood–brain barrier penetration
addition of levofloxacin for the first 2 months of treat‑ of antitubercular drugs is particularly pertinent when
ment10. A probable explanation is that the rifampicin dose choosing the fourth drug of a regimen for drug-sensi‑
was too low to substantially improve CSF drug exposure tive TBM or for constructing a multidrug regimen for
and bacterial killing, and higher doses might yet prove drug-resistant TBM (TABLE 3). The addition of a fourth drug
beneficial. Much higher doses of rifampicin (35 mg/kg) (usually ethambutol) to the intensive phase of pulmo‑
have now been tested in clinical trials for the treatment of nary tuberculosis treatment is primarily to overcome
pulmonary tuberculosis, and resulted in increased rates the possibility of isoniazid resistance; if the bacteria are
of sputum culture conversion by 2 months of treatment123. susceptible to isoniazid then the addition of ethambutol

10 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

Table 3 | Antituberculosis drugs used in TBM treatment

Drug WHO recommended daily dose WHO CSF penetration Important adverse effects
recommended (CSF/plasma
Children Adults duration concentration)
First-line drugs for the treatment of drug-sensitive TBM
Rifampicin 15 mg/kg (range 10 mg/kg (range 12 months 10–20% Hepatotoxicity, orange urine, many drug
10–20 mg/kg); 8–12 mg/kg); interactions
max. 600 mg max. 600 mg
Isoniazid 10 mg/kg (range 5 mg/kg (range 12 months 80–90% Hepatotoxicity, peripheral neuropathy,
7–15 mg/kg; 4–6 mg/kg); max. lupus-like syndrome, confusion, seizures
max. 300 mg 300 mg
Pyrazinamide 35 mg/kg (range 25 mg/kg (range First 2 months 90–100% Hepatotoxicity, arthralgia, gout
30–40 mg/kg) 20–30 mg/kg)
Ethambutol 20 mg/kg (range 15 mg/kg (range First 2 months 20–30% Dose-related retrobulbar neuritis, more
15–25 mg/kg) 15–20 mg/kg) common in renal impairment
Streptomycin* 15–30 mg/kg; 15 mg/kg (range First 2 months 10–20% Monitor plasma concentrations when possible;
max. 1 g IV or IM 12–18 mg/kg); causes nephrotoxicity and ototoxicity
max. 1 g
Core second-line drugs for the treatment of MDR-TBM‡
Levofloxacin <5 years: 10–15 mg/kg Throughout 70–80% Nausea, headache, tremor, confusion, tendon
15–20 mg/kg treatment rupture (rare)
≥5 years:
10–15 mg/kg
Moxifloxacin 10–20 mg/kg: 400 mg Throughout 70–80% Nausea, headache, tremor, confusion, tendon
max. 400 mg treatment rupture (rare)
Not well
established
Amikacin 15–30 mg/kg; 15 mg/kg; max. Intensive phase 10–20% Monitor plasma concentrations when possible;
max. 1 g IV or IM 1 g IV or IM only causes nephrotoxicity and ototoxicity
Kanamycin 15–30 mg/kg; 15 mg/kg; max. Intensive phase 10–20% Monitor plasma concentrations when possible;
max. 1 g IV or IM 1 g IV or IM only causes nephrotoxicity and ototoxicity
Capreomycin 15–30 mg/kg; 15 mg/kg; max. Intensive phase No data (probably Monitor plasma concentrations when possible;
max. 1 g IV or IM 1 g IV or IM only very low) causes nephrotoxicity and ototoxicity
Ethionamide or 15–20 mg/kg; 15–20 mg/kg; Throughout 80–90% Anorexia, nausea, vomiting, gynaecomastia,
prothionamide max. 1 g max. 1 g treatment hypothyroidism, confusion, seizures
Cycloserine 10–20 mg/kg; 10–15 mg/kg; Throughout 80–90% CNS toxicity: depression, seizures,
max. 1 g max. 1 g treatment neuropathy; co‑administration with pyridoxine
recommended
Linezolid 10 mg/kg; max. 600 mg Throughout 30–70% Myelosuppression, optic neuropathy;
600 mg treatment co‑administration with pyridoxine
recommended
Other drugs used to treat MDR tuberculosis, but of uncertain benefit in TBM
Clofazimine 1 mg/kg 100–200 mg No recommended Limited data Can discolour skin orange/red;
duration (probably low) photosensitivity
Para-aminosalicylic 200–300 mg/kg 8–12 g No recommended No data (probably Vomiting, diarrhoea, reversible hypothyroidism
acid duration very low) (increased risk with ethionamide)
Bedaquiline Not determined 400 mg for New drug; limited Probably very low Nausea, vomiting, arthralgia, QT prolongation;
2 weeks, then availability (but data from metabolized through CYP3A4; rifampicin
200 mg thrice one patient only) halves concentrations
weekly
Delamanid Not determined 200 mg New drug; limited No data Nausea, vomiting, dizziness rarely; QT
availability prolongation
*Streptomycin is no longer recommended in the standard regimen for drug-susceptible tuberculosis. ‡WHO recommend an initial regimen with at least five
effective drugs, including pyrazinamide and four second-line drugs. An injectable agent should be used in the intensive phase, the duration of which is uncertain,
but 2011 WHO guidelines recommend at least 8 months. Total duration of treatment for MDR-TBM is uncertain, but should probably be at least 18 months. CSF,
cerebrospinal fluid; IM, intramuscular; IV, intravenous; MDR, multidrug-resistant; TBM, tuberculous meningitis.

does not enhance bacterial killing or improve outcomes. the poor penetration of rifampicin into the brain and
By contrast, in TBM treatment, the rationale for adding CSF, addition of a fourth drug in TBM might also have
a fourth drug is that it should overcome the potential a larger role in overall bacterial killing and increased
adverse effects of isoniazid resistance. However, given patient survival.

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 11

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

No comparative controlled trials have been con‑ linezolid; pyrazinamide and high-dose isoniazid, which
ducted that define the optimal fourth drug for TBM. are often added to second-line treatment, also have good
Ethambutol is widely recommended, but concentrations penetration into the CSF.
in the CSF are <20% of those in the plasma owing to
its poor penetration of the blood–brain barrier, and its Adjunctive host-directed therapies. Intracerebral
contribution to outcome is potentially limited, even in inflammation has long been recognized as an impor‑
the presence of isoniazid resistance117. Historically, strep‑ tant determinant of TBM outcome. The hypothesis that
tomycin has been used widely, but it also penetrates the adjunctive corticosteroids reduce inflammation and
blood–brain barrier poorly — especially once inflamma‑ thereby improve outcome from TBM was first postu‑
tion has subsided — and this property, combined with lated in the early 1950s127, but it took nearly 50 years for
frequent vestibular toxicity (30%) and a high prevalence sufficient randomized controlled trial evidence to accu‑
of resistance, limits its use. Ethionamide passes across mulate before this treatment was widely recommended
the blood–brain barrier more effectively than ethambu‑ in guidelines. A 2016 Cochrane systematic review and
tol or streptomycin, enabling high CSF concentrations meta-analysis of all relevant published trials concluded
to be achieved, and has become part of standard TBM that corticosteroids increase survival in HIV-1-negative
therapy at some centres120. The fluoroquinolones moxi‑ children and adults with TBM, but that the benefit of
floxacin and levofloxacin both have good CSF penetra‑ corticosteroids was uncertain in individuals co‑infected
tion, and high activity against drug-susceptible bacteria with HIV‑1, and was not found to reduce long-term
as well as many drug-resistant bacteria124,125. In a 2016 neurological disability in any group128.
trial of intensified antituberculosis therapy in Vietnam10, Inflammatory complications can occur weeks or even
levofloxacin failed to improve survival when used as a months after the start of antituberculosis chemotherapy
fifth drug alongside high-dose rifampicin and ethambu‑ for TBM, when initial corticosteroid treatment has
tol. However, if the disease was caused by bacteria resist‑ been reduced or stopped129,130. These events are termed
ant to isoniazid, intensified treatment was associated with ‘paradoxical reactions’, or IRIS in individuals started
significantly increased survival. on antiretroviral therapy for HIV co‑infection within
Current evidence suggests that the choice of fourth the past month. Symptoms include fever, progressively
drug in antituberculosis treatment has little influence on worsening headache, and sometimes seizures and reduced
outcome if the disease is caused by bacteria susceptible consciousness. Brain imaging is important to deter‑
to all first-line drugs, although the limited available data mine whether the cause is hydrocephalus, infarction or
mandate caution in forming conclusions. However, in dis‑ tuberculoma formation.
ease caused by isoniazid-resistant bacteria, clinical trial Tuberculomas are the best described paradoxical
data indicate that a high dose (minimum 15 mg/kg) of reactions to TBM treatment, and most are treated with
rifampicin and the addition of levofloxacin as a fifth agent high-dose adjunctive corticosteroids, despite the absence
improve survival. Observational data also indicate that of controlled trial evidence supporting this approach.
the negative effect of isoniazid resistance on survival could Anecdotal evidence suggests that corticosteroids reduce
be reduced by prescribing pyrazinamide throughout symptoms and inflammation in around 50% of patients
treatment (9–12 months)20. with TBM. However, in some cases corticosteroids have
The effect of multidrug resistance to rifampicin and no effect, and symptoms of TBM will persist or worsen.
isoniazid on TBM outcome is catastrophic, with mor‑ In these circumstances, alternative anti-inflammatory
tality >80%20,71,126. This finding is a particular concern agents have been tried, particularly when the tubercu‑
given the global rise in multidrug-resistant tuberculosis, loma involves the optic chiasm or when optochiasmatic
which now stands at 480,000 cases annually4. The main arachnoiditis threatens vision. Some case series suggest
reason for mortality is the inability for drug resistance that thalidomide could help to alleviate these problems,
to be detected sufficiently quickly to enable changes especially when corticosteroids are ineffective131,132. Other
to be made to the antitubercular treatment regimen. studies have reported success with anti-TNF biological
GeneXpert MTB/RIF has enabled same-day detection agents (such as infliximab)133 and IFNγ treatment134.
of rifampicin-resistant bacteria, which can facilitate Brain infarction is the most common cause of long-
the timely switch to second-line agents in some cases. term neurological disability owing to TBM, and is not
However, even in settings in which GeneXpert is avail‑ prevented by corticosteroid treatment44. Two small trials
able, outcomes remain poor owing to the limited sensi‑ have suggested that treatment with aspirin is safe when
tivity of this assay in CSF samples (50–60%), the lack of added to dexamethasone, and might improve outcome
timely susceptibility information on companion drugs, from TBM by reducing the incidence of brain infarc‑
and the unknown brain penetration and efficacy of tion135,136. One of these studies, conducted in India,
second-line drugs for TBM. randomly assigned 118 adults with TBM to standard
The best regimen for the treatment of multidrug- antituberculosis chemotherapy, with or without aspi‑
resistant TBM is unknown, but until additional data rin (150 mg daily), and found a trend towards reduced
become available, the regimens advised by the WHO for stroke incidence with aspirin (24%, compared with 43%
the treatment of multidrug-resistant pulmonary tuber‑ in the control group) and a significant reduction in mor‑
culosis should be used (TABLE 3). Notably, the second-line tality (21.7% versus 43.0%, P = 0.02)135. A second study in
drugs known to have good penetration into the CSF South Africa randomly assigned 146 children with TBM
include fluoroquinolones, ethionamide, cycloserine and to standard antituberculosis chemotherapy plus placebo

12 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

(n = 50), low-dose aspirin (75 mg per 24 h; n = 47), or Vascular compromise is a common complication
high-dose aspirin (100 mg/kg per 24 h; n = 49). In this of TBM, and leads to ischaemia, infarction and raised
trial, aspirin had no significant effect on patient survival, ICP. The techniques available for real-time monitoring
but on‑treatment new hemiplegia was less common in of vascular abnormalities and their effects on cerebral
the high-dose aspirin group136. haemodynamics, ischaemia and ICP are limited, and
Aside from corticosteroids, antiretroviral drugs are are mostly unavailable outside large, well-resourced cen‑
the only adjunctive agents with proven benefit in the tres. The most studied technique is transcranial Doppler
treatment of TBM. Before the advent of antiretrovi‑ ultrasonography, which can provide accurate nonin­
ral therapy, 9‑month mortality in patients with HIV- vasive information on intracranial blood flow velocities
associated TBM in Vietnam was around 65%72; with and identifies areas of intracranial stenosis in the major
antiretroviral treatment, mortality is now <40%10. The vessels in the circle of Willis. Transcranial Doppler can
one remaining area of uncertainty is whether, in anti­ be performed at the bedside, making it a potentially
retroviral therapy-naive patients, this therapy should valuable tool in critically ill and ventilated patients. To
be started concurrently with antituberculosis drugs or date, studies have suggested that transcranial Doppler
delayed until the tuberculosis has been partially or com‑ can reliably diagnose and track TBM-related intracranial
pletely treated. The balance of risks lies between reduc‑ vasculopathy142,143, but is not a reliable indicator of raised
ing the incidence of other potentially fatal opportunistic ICP in children with tuberculous hydrocephalus144.
infections that can result from early initiation of antiret‑ Published data regarding invasive monitoring of
roviral treatment and reducing the risk of IRIS reactions brain tissue oxygenation and ICP are limited. Some
that could result from delayed antiretroviral therapy and centres advocate immediate external ventricular drains
complicate treatment. Trials in pulmonary tuberculosis with pressure monitoring in children who have severe
have shown that early antiretroviral therapy is associ‑ TBM with hydrocephalus, and parenchymal ICP moni‑
ated with increased survival, especially in patients with toring in those without hydrocephalus140, but the role of
very low CD4+ T‑cell counts137,138; conversely, a single these techniques in treatment decision-making remains
trial in adults with TBM comparing early with 2‑month uncertain. One study used a cerebral oxygenation tissue
delayed antiretroviral therapy showed no difference in probe in two children with severe TBM who demon‑
survival139. Grade 4 adverse events were more frequent strated delayed and worsening cerebral ischaemia
in patients given early antiretroviral therapy, suggest‑ despite full conventional therapy and controlled ICP145.
ing that the deferral of antiretroviral therapy for up to These data confirm that the vascular involvement in
2 months could be advantageous in the management of TBM is potentially progressive. Cerebral oxygenation
TBM with an HIV co‑infection. No data are currently monitoring successfully reversed a precipitous decline
available that can inform decisions about antiretro‑ in oxygen readings in one patient, which might have
viral therapy use in children with HIV and TBM, and prevented infarction, indicating a possible benefit for
the risk:benefit analysis is different from that in adults this intervention in selected patients with severe disease.
owing to the high mortality related to untreated HIV in Mannitol has been the long-standing treatment for
young children. raised ICP in traumatic brain injury, although evidence
from the past decade suggests that hypertonic saline is
Supportive management. Raised intracranial pressure equally effective and safer, especially if ICP monitoring is
(ICP), cerebral ischaemia, hydrocephalus, hyponatrae‑ unavailable146,147. Furthermore, hypertonic saline might
mia and seizures all contribute to poor outcome in be effective in treating TBM-associated hyponatraemia,
patients with TBM, and should be actively tested for a common complication of severe disease.
and treated. However, the evidence defining optimal Hyponatraemia (plasma sodium <135 mmol/l)
management of these complications is scant, and physi‑ occurs in around 40–50% of patients with TBM 28.
cians often overlook this important component of TBM Management difficulties stem from the incompletely
care140. Unfortunately, no standardized protocols exist, characterized pathogenesis of hyponatraemia, com‑
and practice varies substantially between centres. bined with few data detailing its optimal treatment.
Cerebral perfusion and oxygenation can be com‑ SIADH (syndrome of inappropriate antidiuretic hor‑
promised globally by hypovolaemia and hypoxia, or mone secretion) and cerebral salt wasting, which possi‑
locally by raised ICP and infection-related vasculitis and bly occur owing to atrial natriuretic peptide secretion148,
infarction. Hydrocephalus is the most common cause are conventionally thought to be the causes of hypo­
of raised ICP, but ICP can also be elevated by cerebral natraemia in TBM, with results from a 2016 study con‑
vasogenic or cytotoxic oedema, a loss of cerebral pres‑ ducted in India suggesting that cerebral salt wasting
sure autoregulation and vascular reactivity, hypercapnia is the most common cause28. However, distinguishing
or hypocapnia, hyponatraemia, and high temperature. between these two conditions can be difficult. By con‑
Thus, clinical management should be directed at the vention, SIADH is managed by fluid restriction, and
careful and continuous monitoring and correction of cerebral salt wasting by fluid administration. Some
abnormalities in gas exchange and tissue oxygenation, have argued that both conditions can be treated with
through mechanical ventilation (if necessary), metic‑ hypertonic saline149, whereas others have argued that
ulous fluid and electrolyte management, control and fluid restriction, the traditional treatment for SIADH,
monitoring (when possible) of intracranial pressure, has little benefit in meningitis and might result in
and rigorous temperature control140,141. worsening hypovolaemia and harm150.

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 13

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Hydrocephalus is the most common cause of raised HIV, substantially reducing the risk of TBM irrespective
ICP. In the majority (80%) of patients with TBM, com‑ of baseline CD4+ T-cell count, tuberculin skin test status,
municating hydrocephalus is observed, caused by and antimycobacterial drug resistance165.
exudates in the basal cisterns and resulting in disrup‑
tion of CSF flow. Reasonable evidence suggests that Conclusions
hydrocephalus in TBM can be treated medically with Progress has been made in understanding the patho‑
diuretics and with repeated lumbar punctures151,152. genesis of TBM, but has not yet translated into outcome
Noncommunicating hydrocephalus, caused by obstruc‑ benefits in patients. Inconsistency between studies has
tion at the level of the cerebral aqueduct and/or the out‑ reduced their generalizability and comparability. A
let foramina of the fourth ventricle, is less common in standardized definition of TBM, which has been pro‑
this population, and should be treated with ventriculo­ posed for use in research studies114, and standardized
peritoneal shunting (VPS) or endoscopic third ventricu‑ methods for enhanced quality and comparability of
lostomy (ETV)153. Conventional CT imaging cannot TBM studies166 will help to increase generalizability and
reliably determine the physical position of the CSF comparability of studies.
block154, and air encephalography has been suggested as High-resolution MRI or CT imaging combined with
a gold standard155. Air encephalography is carried out at imaging techniques to assess TBM disease activity, as
the time of lumbar puncture, with the patient in the sit‑ demonstrated in other forms of tuberculosis167, could
ting position: 10 ml of air are injected into the subarach‑ provide a more detailed clinical phenotype for TBM
noid space, and its journey to the ventricles is shown by than is currently available. This assessment could also
plain skull radiographs 30 min later. Air trapped below be extended to include monitoring of parameters such
the ventricles indicates the presence of noncommuni‑ as brain oxygenation and CSF pressure. Unbiased appli‑
cating hydrocephalus140. However, air encephalography cation of ‘omics’ technologies (particularly transcrip‑
involves a lumbar puncture and carries the risk of cere‑ tomics) has also provided important insight into human
bral herniation; therefore, further exploration of alterna‑ tuberculosis168. Analysis of cells and fluid arising from
tive approaches, such as thin-sliced CT of the posterior the site of disease is likely to be particularly valuable.
fossa or MRI CSF-flow studies, is desirable, although One small autopsy study compared gene transcripts
access to such imaging can be limited in many TBM in brain biopsy samples from patients with TBM and
prevalent settings. patients with traumatic brain injury, and showed that
A 2017 systematic review of VPS in TBM concluded genes encoding proteins involved in metabolism, cell
that the outcome depends on the clinical severity of growth, transport, immune response, cell communica‑
TBM. Patients with an HIV‑1 co‑infection have a worse tion and signal transduction are differentially expressed
prognosis than HIV negative patients156. The main chal‑ in these patient populations169. In addition, previous
lenge for surgeons is the selection of patients for VPS work has shown that longitudinal sampling of blood or
insertion or ETV157,158. One randomized controlled trial CSF during observational studies or randomized con‑
comparing VPS with ETV in 48 children with TBM and trolled trials is highly informative regarding the dynam‑
hydrocephalus suggested that the risk of early recur‑ ics of the disease, such as disease progression and drug
rence of hydrocephalus was higher with ETV, but ETV response31.
had fewer long-term complications than VPS159. Many Delayed diagnosis and treatment is associated with a
studies attest to the high complication rate of VPS160, poor prognosis in TBM. Thus, new tests that can be used
especially in individuals co‑infected with HIV 161. at the point of care are crucially needed, particularly for
However, whether EVT offers a meaningful advantage drug-resistant TBM, which has a dismal outcome. Next-
remains controversial. ETV in TBM hydrocephalus generation nucleic amplification tests might address this
can be technically very difficult, especially in the acute issue. High quality, well-designed head‑to‑head compar‑
stage of disease, owing to the presence of an inflamed, isons that are compliant with Consolidated Standards
thick and opaque third ventricle floor162,163. Others of Reporting Trials (CONSORT) and Standards for
have opined that ETV should be reserved for early-stage Reporting of Diagnostic Accuracy (STARD) are
TBM with aqueductal stenosis, whereas chronic required, as small-scale studies without validation are
burnt-out cases or cases with communicating hydro‑ insufficient to establish a new diagnostic test. The use
cephalus should be managed by ventriculoperitoneal of the absolute gold standard of microbiological confir‑
shunting164. mation is likely to lead to overestimates of diagnostic
sensitivity: further use of composite gold standards and
Prevention latent class analysis, as well as rational decision trees that
The role of BCG in reducing the incidence of dissemi‑ combine diagnostic information with clinical outcome
nated tuberculosis, including TBM, is widely accepted. data, should also form part of the evaluation. In addi‑
Childhood disease can also be prevented by giving tion, improved diagnosis has the potential to improve
isoniazid, or other preventive therapy regimes, to chil‑ national and international surveillance of numbers of
dren exposed to infectious adults. Improvements in cases and trends in incidence.
worldwide tuberculosis control through increased Few studies have rigorously evaluated antibiotic
and more-rapid detection of infection would improve treatment regimens in TBM. Culture positivity was
preven­tion. Antiretroviral therapy is the most effective associated with worse TBM outcome in a large cohort
pre­ventive strategy against tuberculosis in people with of HIV-negative patients and in patients with TBM-IRIS

14 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

in South Africa40; consequently, more-rapid reduction Of 17 clinical trials registered on ClinicalTrials.gov

of bacillary load represents an intuitive approach. under the terms ‘tuberculosis’ or ‘tuberculous’ and
Disappointingly, however, the largest trial of an intensi‑ ‘meningitis’, only two (both of aspirin or corticos‑
fied treatment regimen showed no benefit10. This finding teroid adjunctive therapies) have clinical end points.
emphasizes the need for further detailed pharmaco­ This observation emphasizes the need for biomarkers
dynamic and pharmacokinetic studies in TBM trials of clinical response that might facilitate shorter trials.
so that site‑of‑action pharmacokinetics and exposure– TBM would also be a potential target for multi-arm
response relationships can be explored and regimens multistage clinical trial designs, given the frequently
can be built, with ineffective drugs and doses confi‑ poor outcome of the disease170. Multinational collabo‑
dently discarded. No high-quality studies on the treat‑ rating consortia might go some way towards increasing
ment of drug-resistant TBM are currently available, but advocacy for TBM and enabling high-quality studies
the urgent need for these studies can now be addressed to be undertaken. Further high-quality studies into
with the increasing application of molecular diagnostics optimal supportive treatment for TBM would also be
such as Gene Xpert MTB/RIF testing. desirable.

1. Ducomble, T. et al. The burden of extrapulmonary and 21. Pepper, D. J. et al. Neurologic manifestations of 37. Schoeman, J. F. et al. Adjunctive thalidomide therapy
meningitis tuberculosis: an investigation of national paradoxical tuberculosis-associated immune for childhood tuberculous meningitis: results of a
surveillance data, Germany, 2002 to 2009. Euro reconstitution inflammatory syndrome: a case series. randomized study. J. Child Neurol. 19, 250–257
Surveill. 18, 20436 (2013). Clin. Infect. Dis. 48, e96–e107 (2009). (2004).
2. Pan, Y. et al. Host and microbial predictors of 22. Burn, C. G. & Finley, K. H. The role of hypersensitivity 38. Yadav, A. et al. Correlation of CSF proinflammatory
childhood extrathoracic tuberculosis and tuberculosis in the production of experimental meningitis: I. cytokines with MRI in tuberculous meningitis.
meningitis. Pediatr. Infect. Dis. J. 34, 1289–1295 Experimental meningitis in tuberculous animals. Acad. Radiol. 17, 194–200 (2010).
(2015). J. Exp. Med. 56, 203–221 (1932). 39. Misra, U. K. et al. A study of cytokines in tuberculous
3. Trunz, B. B., Fine, P. & Dye, C. Effect of BCG 23. Peterson, P. K. et al. CD14 receptor-mediated uptake meningitis: clinical and MRI correlation. Neurosci.
vaccination on childhood tuberculous meningitis and of nonopsonized Mycobacterium tuberculosis by Lett. 483, 6–10 (2010).
miliary tuberculosis worldwide: a meta-analysis and human microglia. Infect. Immun. 63, 1598–1602 40. Marais, S. et al. Frequency, severity, and prediction of
assessment of cost-effectiveness. Lancet 367, (1995). tuberculous meningitis immune reconstitution
1173–1180 (2006). 24. Lee, H. M., Kang, J., Lee, S. J. & Jo, E. K. Microglial inflammatory syndrome. Clin. Infect. Dis. 56, 450–460
4. World Health Organization. Global Tuberculosis activation of the NLRP3 inflammasome by the priming (2013).
Report 21st edition (WHO, 2016). signals derived from macrophages infected with 41. Marais, S. et al. Neutrophil-associated central nervous
5. Gomes, T. et al. Epidemiology of extrapulmonary mycobacteria. Glia 61, 441–452 (2013). system inflammation in tuberculous meningitis
tuberculosis in Brazil: a hierarchical model. BMC 25. Ray, G., Aneja, S., Jain, M. & Batra, S. Evaluation of immune reconstitution inflammatory syndrome.
Infect. Dis. 14, 9 (2014). free radical status in CSF in childhood meningitis. Clin. Infect. Dis. 59, 1638–1647 (2014).
6. Chiang, S. S. et al. Treatment outcomes of childhood Ann. Trop. Paediatr. 20, 115–120 (2000). 42. Marais, S. et al. Inflammasome activation underlies
tuberculous meningitis: a systematic review and meta- 26. Mason, S. et al. A hypothetical astrocyte–microglia central nervous system deterioration in HIV-associated
analysis. Lancet Infect. Dis. 14, 947–957 (2014). lactate shuttle derived from a 1H NMR metabolomics tuberculosis. J. Infect. Dis. 215, 677–686 (2016).
7. Marais, S., Pepper, D. J., Schutz, C., Wilkinson, R. J. & analysis of cerebrospinal fluid from a cohort of South 43. Kalita, J., Prasad, S. & Misra, U. K. Predictors of
Meintjes, G. Presentation and outcome of tuberculous African children with tuberculous meningitis. paradoxical tuberculoma in tuberculous meningitis.
meningitis in a high HIV prevalence setting. PLoS ONE Metabolomics 11, 822–837 (2015). Int. J. Tuberc. Lung Dis. 18, 486–491 (2014).
6, e20077 (2011). 27. Thwaites, G. E. et al. Pathophysiology and prognosis in 44. Thwaites, G. E. et al. Serial MRI to determine the
8. Rana, F. S. et al. Autopsy study of HIV‑1‑positive and vietnamese adults with tuberculous meningitis. effect of dexamethasone on the cerebral pathology of
HIV‑1‑negative adult medical patients in Nairobi. J. Infect. Dis. 188, 1105–1115 (2003). tuberculous meningitis: an observational study.
Kenya. J. Acquir. Immune Def. Syndr. 24, 23–29 28. Misra, U. K., Kalita, J., Bhoi, S. K. & Singh, R. K. A Lancet Neurol. 6, 230–236 (2007).
(2000). study of hyponatremia in tuberculous meningitis. 45. Thomas, M. M. et al. Rapid diagnosis of
9. Thwaites, G. E. et al. The influence of HIV infection on J. Neurol. Sci. 367, 152–157 (2016). Mycobacterium tuberculosis meningitis by
clinical presentation, response to treatment, and 29. Dhanwal, D. K., Vyas, A., Sharma, A. & Saxena, A. enumeration of cerebrospinal fluid antigen-specific
outcome in adults with Tuberculous meningitis. Hypothalamic pituitary abnormalities in tubercular T‑cells. Int. J. Tuberc. Lung Dis. 12, 651–657 (2008).
J. Infect. Dis. 192, 2134–2141 (2005). meningitis at the time of diagnosis. Pituitary 13, 46. Caccamo, N. et al. Phenotypical and functional
10. Heemskerk, A. D. et al. Intensified antituberculosis 304–310 (2010). analysis of memory and effector human CD8 T cells
therapy in adults with tuberculous meningitis. 30. Rock, R. B. et al. Mycobacterium tuberculosis-induced specific for mycobacterial antigens. J. Immunol. 177,
N. Engl. J. Med. 374, 124–134 (2016). cytokine and chemokine expression by human 1780–1785 (2006).
11. Tho, D. Q. et al. Influence of antituberculosis drug microglia and astrocytes: effects of dexamethasone. 47. Dieli, F. et al. Predominance of Vγ9/Vδ2 T lymphocytes
resistance and Mycobacterium tuberculosis lineage J. Infect. Dis. 192, 2054–2058 (2005). in the cerebrospinal fluid of children with tuberculous
on outcome in HIV-associated tuberculous meningitis. 31. Simmons, C. P. et al. The clinical benefit of adjunctive meningitis: reversal after chemotherapy. Mol. Med. 5,
Antimicrob. Agents Chemother. 56, 3074–3079 (2012). dexamethasone in tuberculous meningitis is not 301–312 (1999).
12. Rich, A. R. The Pathogenesis of Tuberculosis associated with measurable attenuation of peripheral 48. Mansour, A. M. et al. Relationship between
(C. C. Thomas, 1946). or local immune responses. J. Immunol. 175, intracranial granulomas and cerebrospinal fluid levels
13. Donald, P. R. & Schoeman, J. F. Tuberculous 579–590 (2005). of gamma interferon and interleukin‑10 in patients
meningitis. N. Engl. J. Med. 351, 1719–1720 (2004). 32. Simmons, C. P. et al. Pretreatment intracerebral and with tuberculous meningitis. Clin. Diagn. Lab Immunol.
14. Tucker, E. W. et al. Microglia activation in a pediatric peripheral blood immune responses in vietnamese 12, 363–365 (2005).
rabbit model of tuberculous meningitis. Dis. Model. adults with tuberculous meningitis: diagnostic value 49. Yang, Q. et al. IP‑10 and MIG are compartmentalized
Mech. 9, 1497–1506 (2016). and relationship to disease severity and outcome. at the site of disease during pleural and meningeal
15. van Leeuwen, L. M. et al. Modeling tuberculous J. Immunol. 176, 2007–2014 (2006). tuberculosis and are decreased after antituberculosis
meningitis in zebrafish using Mycobacterium 33. Tsenova, L., Sokol, K., Freedman, V. H. & Kaplan, G. A treatment. Clin. Vaccine Immunol. 21, 1635–1644
marinum. Dis. Model. Mech. 7, 1111–1122 (2014). combination of thalidomide plus antibiotics protects (2014).
16. Donald, P. R., Schaaf, H. S. & Schoeman, J. F. rabbits from mycobacterial meningitis-associated 50. Green, J. A. et al. Dexamethasone, cerebrospinal fluid
Tuberculous meningitis and miliary tuberculosis: the death. J. Infect. Dis. 177, 1563–1572 (1998). matrix metalloproteinase concentrations and clinical
Rich focus revisited. J. Infect. 50, 193–195 (2005). 34. Tsenova, L., Bergtold, A., Freedman, V. H., outcomes in tuberculous meningitis. PLoS ONE 4,
17. Berenguer, J. et al. Tuberculous meningitis in patients Young, R. A. & Kaplan, G. Tumor necrosis factor α is a e7277 (2009).
infected with the human immunodeficiency virus. determinant of pathogenesis and disease progression 51. Green, J. A. et al. Mycobacterium tuberculosis-
N. Engl. J. Med. 326, 668–672 (1992). in mycobacterial infection in the central nervous infected human monocytes down-regulate microglial
18. Shafer, R. W., Goldberg, R., Sierra, M. & Glatt, A. E. system. Proc. Natl Acad. Sci. USA 96, 5657–5662 MMP‑2 secretion in CNS tuberculosis via TNFα, NFκB,
Frequency of Mycobacterium tuberculosis bacteremia (1999). p38 and caspase 8 dependent pathways.
in patients with tuberculosis in an area endemic for 35. Akalin, H., Akdis, A. C., Mistik, R., Helvaci, S. & J. Neuroinflamm. 8, 46 (2011).
AIDS. Am. Rev. Respir. Dis. 140, 1611–1613 (1989). Kilicturgay, K. Cerebrospinal fluid interleukin‑1ß/ 52. Visser, D. H. et al. Host immune response to
19. Marais, S., Pepper, D. J., Marais, B. J. & Torok, M. E. interleukin‑1 receptor antagonist balance and tumor tuberculous meningitis. Clin. Infect. Dis. 60, 177–187
HIV-associated tuberculous meningitis — diagnostic necrosis factor-α concentrations in tuberculous, viral (2015).
and therapeutic challenges. Tuberculosis 90, 367–374 and acute bacterial meningitis. Scand. J. Infect. Dis. 53. Matsuyama, W. et al. Expression of vascular
(2010). 26, 667–674 (1994). endothelial growth factor in tuberculous meningitis.
20. Thwaites, G. E. et al. Effect of antituberculosis drug 36. Donald, P. R. et al. Concentrations of interferon γ, J. Neurol. Sci. 186, 75–79 (2001).
resistance on response to treatment and outcome in tumor necrosis factor α, and interleukin‑1ß in the 54. Misra, U. K., Kalita, J., Singh, A. P. & Prasad, S.
adults with tuberculous meningitis. J. Infect. Dis. 192, cerebrospinal fluid of children treated for tuberculous Vascular endothelial growth factor in tuberculous
79–88 (2005). meningitis. Clin. Infect. Dis. 21, 924–929 (1995). meningitis. Int. J. Neurosci. 123, 128–132 (2013).

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 15

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

55. Ozden, M., Kalkan, A., Demirdag, K., Denk, A. & 79. Rohlwink, U. K. et al. Imaging features of the brain, meningitis. J. Stroke Cerebrovasc Dis. 18, 251–258
Kilic, S. S. Hepatocyte growth factor (HGF) in patients cerebral vessels and spine in pediatric tuberculous (2009).
with hepatitis B and meningitis. J. Infect. 49, 229–235 meningitis with associated hydrocephalus. Pediatr. 103. Dekker, G. et al. MRI findings in children with
(2004). Infect. Dis. J. 35, e301–e310 (2016). tuberculous meningitis: a comparison of HIV-infected
56. Comas, I. et al. Out‑of‑Africa migration and Neolithic 80. Thwaites, G. E., Chau, T. T. & Farrar, J. J. Improving the and non-infected patients. Childs Nerv. Syst. 27,
coexpansion of Mycobacterium tuberculosis with bacteriological diagnosis of tuberculous meningitis. 1943–1949 (2011).
modern humans. Nat. Genet. 45, 1176–1182 (2013). J. Clin. Microbiol. 42, 378–379 (2004). 104. Andronikou, S. et al. Value of early follow‑up CT in
57. Caws, M. et al. The influence of host and bacterial 81. Stewart, S. M. The bacteriological diagnosis of paediatric tuberculous meningitis. Pediatr. Radiol. 35,
genotype on the development of disseminated disease tuberculous meningitis. J. Clin. Pathol. 6, 241–242 1092–1099 (2005).
with Mycobacterium tuberculosis. PLoS Pathog. 4, (1953). 105. Gambhir, S. et al. Role of 18F‑FDG PET in demonstrating
e1000034 (2008). 82. Caws, M. et al. Evaluation of the MODS culture disease burden in patients with tuberculous
58. Faksri, K. et al. Epidemiological trends and clinical technique for the diagnosis of tuberculous meningitis. meningitis. J. Neurol. Sci. 370, 196–200 (2016).
comparisons of Mycobacterium tuberculosis lineages PLoS ONE 2, e1173 (2007). 106. Solomons, R. S. et al. Chest radiograph findings in
in Thai TB meningitis. Tuberculosis (Edinb.) 91, 83. Pai, M. et al. Diagnostic accuracy of nucleic acid children with tuberculous meningitis. Int. J. Tuberc.
594–600 (2011). amplification tests for tuberculous meningitis: a Lung Dis. 19, 200–204 (2015).
59. Nicol, M. P. et al. Distribution of strain families of systematic review and meta-analysis. Lancet Infect. 107. Thwaites, G. E. et al. Diagnosis of adult tuberculous
Mycobacterium tuberculosis causing pulmonary and Dis. 3, 633–643 (2003). meningitis by use of clinical and laboratory features.
extrapulmonary disease in hospitalized children in 84. Patel, V. B. et al. Diagnostic accuracy of quantitative Lancet 360, 1287–1292 (2002).
Cape Town, South Africa. J. Clin. Microbiol. 43, PCR (Xpert MTB/RIF) for tuberculous meningitis in a 108. Sunbul, M., Atilla, A., Esen, S., Eroglu, C. &
5779–5781 (2005). high burden setting: a prospective study. PLoS Med. Leblebicioglu, H. Thwaites’ diagnostic scoring and the
60. Wang, J. et al. DNA polymorphism of Mycobacterium 10, e1001536 (2013). prediction of tuberculous meningitis. Med. Princ.
tuberculosis PE_PGRS33 gene among clinical isolates 85. Bahr, N. C. et al. Improved diagnostic sensitivity for Pract. 14, 151–154 (2005).
of pediatric TB patients and its associations with tuberculous meningitis with Xpert((R)) MTB/RIF of 109. Torok, M. E. et al. Validation of a diagnostic algorithm
clinical presentation. Tuberculosis (Edinb.) 91, centrifuged CSF. Int. J. Tuberc. Lung Dis. 19, for adult tuberculous meningitis. Am. J. Trop. Med.
287–292 (2011). 1209–1215 (2015). Hyg. 77, 555–559 (2007).
61. Tobin, D. M. et al. Host genotype-specific therapies 86. Nhu, N. T. et al. Evaluation of GeneXpert MTB/RIF for 110. Vibha, D. et al. Validation of diagnostic algorithm to
can optimize the inflammatory response to diagnosis of tuberculous meningitis. J. Clin. Microbiol. differentiate between tuberculous meningitis and
mycobacterial infections. Cell 148, 434–446 (2012). 52, 226–233 (2014). acute bacterial meningitis. Clin. Neurol. Neurosurgery
62. Thuong, N. et al. Leukotriene A4 hydrolase genotype 87. Boyles, T. H. & Thwaites, G. E. Appropriate use of the 114, 639–644 (2012).
and HIV infection influence intracerebral inflammation Xpert(R) MTB/RIF assay in suspected tuberculous 111. Zhang, Y. L., Lin, S., Shao, L. Y., Zhang, W. H. &
and survival from tuberculous meningitis J. Infect. Dis. meningitis. Int. J. Tuberc. Lung Dis. 19, 276–277 Weng, X. H. Validation of Thwaites’ diagnostic scoring
215, 1020–1028 (2017). (2015). system for the differential diagnosis of tuberculous
63. van Laarhoven, A. et al. Clinical parameters, routine 88. Bahr, N. C. et al. GeneXpert MTB/Rif to diagnose meningitis and bacterial meningitis. Jpn. J. Infect. Dis.
inflammatory markers, and LTA4H genotype as tuberculous meningitis: perhaps the first test but not 67, 428–431 (2014).
predictors of mortality among 608 patients with the last. Clin. Infect. Dis. 62, 1133–1135 (2016). 112. Saavedra, J. S. et al. Validation of Thwaites Index for
tuberculous meningitis in Indonesia. J. Infect. Dis. 89. World Health Organization. Meeting Report of a diagnosing tuberculous meningitis in a Colombian
215, 1029–1039 (2017). Technical Expert Consultation: Non-inferiority Analysis population. J. Neurol. Sci. 370, 112–118 (2016).
64. Thwaites, G. E., van Toorn, R. & Schoeman, J. of Xpert MTB/RIF Ultra Compared to Xpert MTB/RIF 113. Checkley, A. M., Njalale, Y., Scarborough, M. &
Tuberculous meningitis: more questions, still too few (WHO, 2017). Zjilstra, E. E. Sensitivity and specificity of an index for
answers. Lancet Neurol. 12, 999–1010 (2013). 90. Cox, J. A. et al. Accuracy of lipoarabinomannan and the diagnosis of TB meningitis in patients in an urban
65. van Well, G. T. et al. Twenty years of pediatric Xpert MTB/RIF testing in cerebrospinal fluid to teaching hospital in Malawi. Trop. Med. Int. Health 13,
tuberculous meningitis: a retrospective cohort study in diagnose tuberculous meningitis in an autopsy cohort 1042–1046 (2008).
the western cape of South Africa. Pediatrics 123, of HIV-infected adults. J. Clin. Microbiol. 53, 114. Marais, S. et al. Tuberculous meningitis: a uniform
e1–e8 (2009). 2667–2673 (2015). case definition for use in clinical research. Lancet
66. Miftode, E. G. et al. Tuberculous meningitis in children 91. Patel, V. B. et al. Utility of a novel lipoarabinomannan Infect. Dis. 10, 803–812 (2010).
and adults: a 10‑year retrospective comparative assay for the diagnosis of tuberculous meningitis in a 115. Solomons, R. S., Visser, D. H., Marais, B. J.,
analysis. PLoS ONE 10, e0133477 (2015). resource-poor high-HIV prevalence setting. Schoeman, J. F. & van Furth, A. M. Diagnostic
67. Torok, M. E. et al. Clinical and microbiological features Cerebrospinal Fluid Res. 6, 13 (2009). accuracy of a uniform research case definition for TBM
of HIV-associated tuberculous meningitis in 92. Patel, V. B. et al. Comparison of a clinical prediction in children: a prospective study. Int. J. Tuberc. Lung
Vietnamese adults. PLoS ONE 3, e1772 (2008). rule and a LAM antigen-detection assay for the rapid Dis. 20, 903–908 (2016).
68. Streptomycin in Tuberculosis Trials Commitee, Medical diagnosis of TBM in a high HIV prevalence setting. 116. Donald, P. R. Cerebrospinal fluid concentrations of
Research Council. Streptomycin treatment of PLoS ONE 5, e15664 (2010). antituberculosis agents in adults and children.
tuberculous meningitis. Lancet 1, 582–596 (1948). 93. Yu, J., Wang, Z. J., Chen, L. H. & Li, H. H. Diagnostic Tuberculosis (Edinb.) 90, 279–292 (2010).
69. Dhawan, S. R. et al. Predictors of neurological accuracy of interferon-gamma release assays for 117. Donald, P. R. The chemotherapy of tuberculous
outcome of tuberculous meningitis in childhood: A tuberculous meningitis: a meta-analysis. Int. J. Tuberc. meningitis in children and adults. Tuberculosis (Edinb.)
prospective cohort study from a developing country. Lung Dis. 20, 494–499 (2016). 90, 375–392 (2010).
J. Child Neurol. 31, 1622–1627 (2016). 94. Tuon, F. F. et al. Adenosine deaminase and tuberculous 118. World Health Organization. Guidance for National
70. Brancusi, F., Farrar, J. & Heemskerk, D. Tuberculous meningitis — a systematic review with meta-analysis. Tuberculosis Programmes on the Management of
meningitis in adults: a review of a decade of Scand. J. Infect. Dis. 42, 198–207 (2010). Tuberculosis in Children (WHO, 2014).
developments focusing on prognostic factors for 95. Xu, H. B., Jiang, R. H., Li, L., Sha, W. & Xiao, H. P. 119. Thwaites, G. et al. British Infection Society guidelines
outcome. Future Microbiol. 7, 1101–1116 (2012). Diagnostic value of adenosine deaminase in for the diagnosis and treatment of tuberculosis of the
71. Vinnard, C. et al. The long-term mortality of cerebrospinal fluid for tuberculous meningitis: a meta- central nervous system in adults and children.
tuberculosis meningitis patients in new york city: a analysis. Int. J. Tuberc. Lung Dis. 14, 1382–1387 J. Infect. 59, 167–187 (2009).
cohort study. Clin. Infect. Dis. 64, 401–407 (2016). (2010). 120. van Toorn, R. et al. Short intensified treatment in
72. Thwaites, G. E. et al. Dexamethasone for the 96. Bullock, M. R. & Welchman, J. M. Diagnostic and children with drug-susceptible tuberculous meningitis.
treatment of tuberculous meningitis in adolescents prognostic features of tuberculous meningitis on CT Pediatr. Infect. Dis. J. 33, 248–252 (2014).
and adults. N. Engl. J. Med. 351, 1741–1751 scanning. J. Neurol. Neurosurg. Psychiatry 45, 121. Savic, R. M. et al. Pediatric tuberculous meningitis:
(2004). 1098–1101 (1982). model-based approach to determining optimal doses
73. Garg, R. K., Malhotra, H. S. & Jain, A. Neuroimaging 97. Andronikou, S., Smith, B., Hatherhill, M., Douis, H. & of the anti-tuberculosis drugs rifampin and
in tuberculous meningitis. Neurol. India 64, 219–227 Wilmshurst, J. Definitive neuroradiological diagnostic levofloxacin for children. Clin. Pharmacol. Ther. 98,
(2016). features of tuberculous meningitis in children. Pediatr. 622–629 (2015).
74. van der Merwe, D. J., Andronikou, S., Van Toorn, R. & Radiol. 34, 876–885 (2004). 122. Ruslami, R. et al. Intensified regimen containing
Pienaar, M. Brainstem ischemic lesions on MRI in 98. Przybojewski, S., Andronikou, S. & Wilmshurst, J. rifampicin and moxifloxacin for tuberculous
children with tuberculous meningitis: with diffusion Objective CT criteria to determine the presence of meningitis: an open-label, randomised controlled
weighted confirmation. Childs Nerv. Syst. 25, abnormal basal enhancement in children with phase 2 trial. Lancet Infect. Dis. 13, 27–35 (2013).
949–954 (2009). suspected tuberculous meningitis. Pediatr. Radiol 36, 123. Boeree, M. J. et al. High-dose rifampicin, moxifloxacin,
75. Omar, N., Andronikou, S., van Toorn, R. & Pienaar, M. 687–696 (2006). and SQ109 for treating tuberculosis: a multi-arm,
Diffusion-weighted magnetic resonance imaging of 99. Botha, H. et al. Reliability and diagnostic performance multi-stage randomised controlled trial. Lancet Infect.
borderzone necrosis in paediatric tuberculous of CT imaging criteria in the diagnosis of tuberculous Dis. 17, 39–49 (2017).
meningitis. J. Med. Imag. Radiat. Oncol. 55, 563–570 meningitis. PLoS ONE 7, e38982 (2012). 124. Alffenaar, J. W. et al. Pharmacokinetics of moxifloxacin
(2011). 100. Pienaar, M., Andronikou, S. & van Toorn, R. MRI to in cerebrospinal fluid and plasma in patients with
76. Singh, B. et al. Computed tomography angiography in demonstrate diagnostic features and complications of tuberculous meningitis. Clin. Infect. Dis. 49,
patients with tuberculous meningitis. J. Infect. 64, TBM not seen with CT. Childs Nerv. Syst. 25, 1080–1082 (2009).
565–572 (2012). 941–947 (2009). 125. Thwaites, G. E. et al. Randomized pharmacokinetic
77. Lu, T. T. et al. Magnetic resonance angiography 101. Janse van Rensburg, P., Andronikou, S., van Toorn, R. and pharmacodynamic comparison of fluoroquinolones
manifestations and prognostic significance in HIV- & Pienaar, M. Magnetic resonance imaging of miliary for tuberculous meningitis. Antimicrob. Agents
negative tuberculosis meningitis. Int. J. Tuberc. Lung tuberculosis of the central nervous system in children Chemother. 55, 3244–3253 (2011).
Dis. 19, 1448–1454 (2015). with tuberculous meningitis. Pediatr. Radiol 38, 126. Seddon, J. A. et al. Impact of drug resistance on
78. Kalita, J., Prasad, S., Maurya, P. K., Kumar, S. & 1306–1313 (2008). clinical outcome in children with tuberculous
Misra, U. K. MR angiography in tuberculous 102. Kalita, J., Misra, U. K. & Nair, P. P. Predictors of stroke meningitis. Pediatr. Infect. Dis. J. 31, 711–716
meningitis. Acta Radiol. 53, 324–329 (2012). and its significance in the outcome of tuberculous (2012).

16 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

127. Shane, S. J., Clowater, R. A. & Riley, C. The treatment strong is the evidence? Scand. J. Infect. Dis. 33, 175. Thuong, N. T. et al. A polymorphism in human TLR2 is
of tuberculous meningitis with cortisone and 13–26 (2001). associated with increased susceptibility to tuberculous
streptomycin. Can. Med. Assoc. J. 67, 13–15 (1952). 151. Schoeman, J., Donald, P., van Zyl, L., Keet, M. & meningitis. Genes Immun. 8, 422–428 (2007).
128. Prasad, K., Singh, M. B. & Ryan, H. Corticosteroids for Wait, J. Tuberculous hydrocephalus: comparison of 176. Graustein, A. D. et al. TLR9 gene region
managing tuberculous meningitis. Cochrane Database different treatments with regard to ICP, ventricular size polymorphisms and susceptibility to tuberculosis in
Syst. Rev. 4, CD002244 (2016). and clinical outcome. Dev. Med. Child Neurol. 33, Vietnam. Tuberculosis (Edinb.) 95, 190–196 (2015).
129. Singh, A. K. et al. Paradoxical reaction in tuberculous 396–405 (1991). 177. Dissanayeke, S. R. et al. Polymorphic variation in
meningitis: presentation, predictors and impact on 152. Visudhiphan, P. & Chiemchanya, S. Hydrocephalus in TIRAP is not associated with susceptibility to
prognosis. BMC Infect. Dis. 16, 306 (2016). tuberculous meningitis in children: treatment with childhood TB but may determine susceptibility to TBM
130. Garg, R. K., Malhotra, H. S. & Kumar, N. Paradoxical acetazolamide and repeated lumbar puncture. in some ethnic groups. PLoS ONE 4, e6698 (2009).
reaction in HIV negative tuberculous meningitis. J. Pediatr. 95, 657–660 (1979). 178. Hawn, T. R. et al. A polymorphism in toll-interleukin 1
J. Neurol. Sci. 340, 26–36 (2014). 153. Figaji, A. A., Fieggen, A. G. & Peter, J. C. Endoscopic receptor domain containing adaptor protein is
131. Schoeman, J. F., Andronikou, S., Stefan, D. C., third ventriculostomy in tuberculous meningitis. Childs associated with susceptibility to meningeal
Freeman, N. & van Toorn, R. Tuberculous meningitis- Nerv. Syst. 19, 217–225 (2003). tuberculosis. J. Infect. Dis. 194, 1127–1134 (2006).
related optic neuritis: recovery of vision with 154. Bruwer, G. E., Van der Westhuizen, S., Lombard, C. J. 179. Horne, D. J. et al. Common polymorphisms in the
thalidomide in 4 consecutive cases. J. Child Neurol. & Schoeman, J. F. Can CT predict the level of CSF PKP3–SIGIRR–TMEM16J gene region are associated
25, 822–828 (2010). block in tuberculous hydrocephalus? Childs Nerv. Syst. with susceptibility to tuberculosis. J. Infect. Dis. 205,
132. Schoeman, J. F., Fieggen, G., Seller, N., Mendelson, M. 20, 183–187 (2004). 586–594 (2012).
& Hartzenberg, B. Intractable intracranial tuberculous 155. Figaji, A. A., Fieggen, A. G. & Peter, J. C. Re: 180. Feng, W. X. et al. Tag SNP polymorphism of CCL2 and
infection responsive to thalidomide: report of four Endoscopic third ventriculostomy for chronic its role in clinical tuberculosis in Han Chinese pediatric
cases. J. Child Neurol. 21, 301–308 (2006). hydrocephalus after tuberculous meningitis population. PLoS ONE 6, e14652 (2011).
133. Molton, J. S., Huggan, P. J. & Archuleta, S. Infliximab [Jonathan A, Rajshekhar, V Surg Neurol 63 (2005) 181. Kumar, R., Singh, S. N. & Kohli, N. A diagnostic rule
therapy in two cases of severe neurotuberculosis 32–35]. Surg. Neurol. 64, 95 (2005). for tuberculous meningitis. Arch. Dis. Child. 81,
paradoxical reaction. Med. J. Aust. 202, 156–157 156. Rizvi, I. et al. Ventriculo-peritoneal shunt surgery for 221–224 (1999).
(2015). tuberculous meningitis: a systematic review. J. Neurol. 182. Youssef, F. G. et al. Differentiation of tuberculous
134. Lee, J. Y., Yim, J. J. & Yoon, B. W. Adjuvant interferon-γ Sci. 375, 255–263 (2017). meningitis from acute bacterial meningitis using
treatment in two cases of refractory tuberculosis of the 157. Sil, K. & Chatterjee, S. Shunting in tuberculous simple clinical and laboratory parameters. Diagn.
brain. Clin. Neurol. Neurosurgery 114, 732–734 meningitis: a neurosurgeon’s nightmare. Childs Nerv. Microbiol. Infect. Dis. 55, 275–278 (2006).
(2012). Syst. 24, 1029–1032 (2008). 183. Cohen, D. B. et al. Diagnosis of cryptococcal and
135. Misra, U. K., Kalita, J. & Nair, P. P. Role of aspirin in 158. Figaji, A. A. & Fieggen, A. G. Endoscopic challenges tuberculous meningitis in a resource-limited African
tuberculous meningitis: a randomized open label and applications in tuberculous meningitis. World setting. Trop. Med. Int. Health 15, 910–917 (2010).
placebo controlled trial. J. Neurol. Sci. 293, 12–17 Neurosurg. 79, S24e29–S24e14 (2013). 184. Hristea, A. et al. Clinical prediction rule for
(2010). 159. Goyal, P. et al. A randomized study of differentiating tuberculous from viral meningitis.
136. Schoeman, J. F., Janse van Rensburg, A., ventriculoperitoneal shunt versus endoscopic third Int. J. Tuberc. Lung Dis. 16, 793–798 (2012).
Laubscher, J. A. & Springer, P. The role of aspirin in ventriculostomy for the management of tubercular 185. Dendane, T. et al. A simple diagnostic aid for
childhood tuberculous meningitis. J. Child Neurol. 26, meningitis with hydrocephalus. Childs Nerv. Syst. 30, tuberculous meningitis in adults in Morocco by use of
956–962 (2011). 851–857 (2014). clinical and laboratory features. Int. J. Infect. Dis. 17,
137. Havlir, D. V. et al. Timing of antiretroviral therapy for 160. Lamprecht, D., Schoeman, J., Donald, P. & e461–e465 (2013).
HIV‑1 infection and tuberculosis. N. Engl. J. Med. 365, Hartzenberg, H. Ventriculoperitoneal shunting in 186. Zhang, B., Lv, K., Bao, J., Lu, C. & Lu, Z. Clinical and
1482–1491 (2011). childhood tuberculous meningitis. Br. J. Neurosurg. laboratory factors in the differential diagnosis of
138. Blanc, F. X. et al. Earlier versus later start of 15, 119–125 (2001). tuberculous and cryptococcal meningitis in adult HIV-
antiretroviral therapy in HIV-infected adults with 161. Sharma, R. M. et al. Tubercular meningitis with negative patients. Intern. Med. 52, 1573–1578
tuberculosis. N. Engl. J. Med. 365, 1471–1481 hydrocephalus with HIV co‑infection: role of (2013).
(2011). cerebrospinal fluid diversion procedures. J. Neurosurg. 187. Qamar, F. N., Rahman, A. J., Iqbal, S. & Humayun, K.
139. Torok, M. E. et al. Timing of initiation of antiretroviral 122, 1087–1095 (2015). Comparison of clinical and CSF profiles in children
therapy in human immunodeficiency virus (HIV) — 162. Yadav, Y. R. et al. Role of endoscopic third with tuberculous and pyogenic meningitis; role of CSF
associated tuberculous meningitis. Clin. Infect. Dis. ventriculostomy in tuberculous meningitis with protein: glucose ratio as diagnostic marker of
52, 1374–1383 (2011). hydrocephalus. Asian J. Neurosurgery 11, 325–329 tuberculous meningitis. J. Pak. Med. Assoc. 63,
140. Figaji, A. A. & Fieggen, A. G. The neurosurgical and (2016). 206–210 (2013).
acute care management of tuberculous meningitis: 163. Rajshekhar, V. Surgery for brain tuberculosis: a review.
evidence and current practice. Tuberculosis (Edinb.) Acta Neurochir. (Wien) 157, 1665–1678 (2015). Acknowledgements
90, 393–400 (2010). 164. Kumar, A., Singh, K. & Sharma, V. Surgery in Tuberculous Meningitis International Research Consortium
141. Murthy, J. M. Management of intracranial pressure in hydrocephalus of tubercular origin: challenges and members include: Rob Aarnoutse, Reinout Van Crevel,
tuberculous meningitis. Neurocrit. Care 2, 306–312 management. Acta Neurochir. (Wien) 155, 869–873 Aarjan van Laarhoven, Sofiati Dian (Radboud University
(2005). (2013). Nijmegen Medical Center, Nijmegen, The Netherlands);
142. Tai, M. S. & Sharma, V. K. Role of transcranial doppler 165. Lawn, S. D. & Wilkinson, R. J. ART and prevention of Nathan C. Bahr (University of Kansas Medical Center,
in the evaluation of vasculopathy in tuberculous HIV-associated tuberculosis. Lancet HIV 2, e221–e222 Kansas City, USA); David R. Boulware (University of
meningitis. PLoS ONE 11, e0164266 (2016). (2015). Minnesota, Minneapolis, USA); Maxine Caws (Liverpool
143. Kilic, T., Elmaci, I., Ozek, M. M. & Pamir, M. N. Utility 166. Marais, B. J. et al. Standardized methods for School of Tropical Medicine, Liverpool, UK); Mark R. Cronan,
of transcranial Doppler ultrasonography in the enhanced quality and comparability of tuberculous David Tobin (Duke University School of Medicine, Durham,
diagnosis and follow‑up of tuberculous meningitis- meningitis studies. Clin. Infect. Dis. 64, 501–509 USA); Kelly Dooley (Johns Hopkins University School of
related vasculopathy. Childs Nerv. Syst. 18, 142–146 (2017). Medicine, Baltimore, USA); Sarah Dunstan (University of
(2002). 167. Esmail, H. et al. Characterization of progressive HIV- Melbourne, Melbourne, Australia); Guo-dong Feng, Xiaodan
144. van Toorn, R., Schaaf, H. S., Solomons, R., associated tuberculosis using 2‑deoxy‑2-[18F]fluoro- Shi, Ting Wang (Fourth Military Medical University, Xi’an,
Laubscher, J. A. & Schoeman, J. F. The value of D‑glucose positron emission and computed People’s Republic of China); Anthony Figaji, Suzaan Marais,
transcranial Doppler imaging in children with tomography. Nat. Med. 22, 1090–1093 (2016). Helen McIlleron, Graeme Meintjes, Ursula Rohlwink
tuberculous meningitis. Childs Nerv. Syst. 30, 168. Berry, M. P. et al. An interferon-inducible neutrophil- (University of Cape Town, Cape Town, South Africa); Ahmad
1711–1716 (2014). driven blood transcriptional signature in human Rizal, Rovina Ruslami (Padjadjaran University, Bandung,
145. Figaji, A. A. et al. Continuous monitoring and tuberculosis. Nature 466, 973–977 (2010). Indonesia); Ravindra K. Garg (King George Medical
intervention for cerebral ischemia in tuberculous 169. Kumar, G. S. et al. Gene expression profiling of University, Lucknow, India); Mudit Gupta, Rakesh K. Gupta
meningitis. Pediatr. Crit. Care Med. 9, e25–e30 tuberculous meningitis co‑infected with HIV. (Fortis Memorial Research Institute, Gurgaon, India); Sneha
(2008). J. Proteom. Bioinform. 5, 235–244 (2012). Gupta, Rada Savic (University of California, San Francisco,
146. Oddo, M. et al. Effect of mannitol and hypertonic 170. Bratton, D. J., Phillips, P. P. & Parmar, M. K. A USA); Anna D. Heemskerk, Thuong Thuy Thuong Nguyễn,
saline on cerebral oxygenation in patients with severe multi-arm multi-stage clinical trial design for binary Mai Thi Hoàng Nguyễn, Vijay Srinivasan, Guy Thwaites,
traumatic brain injury and refractory intracranial outcomes with application to tuberculosis. BMC Med. Trâm Thi Bích Trân, Thinh Thi Vân Trân, Anh Thi Ngoc Trân,
hypertension. J. Neurol. Neurosurg. Psychiatry 80, Res. Methodol. 13, 139 (2013). Trang Hồng Yêng Võ, Marcel Wolbers (Oxford University
916–920 (2009). 171. Campo, M. et al. Common polymorphisms in the Clinical Research Unit, Ho Chi Minh City, Vietnam); Jayantee
147. Francony, G. et al. Equimolar doses of mannitol and CD43 gene region are associated with tuberculosis Kalita, Usha K. Misra (Sanjay Gandhi Postgraduate Institute
hypertonic saline in the treatment of increased disease and mortality. Am. J. Respir. Cell. Mol. Biol. of Medical Sciences, Lucknow, India); Rachel Lai (The Francis
intracranial pressure. Crit. Care Med. 36, 795–800 52, 342–348 (2015). Crick Institute, London, UK); Ben J. Marais, Mai Quỳnh Trinh
(2008). 172. Hoal-Van Helden, E. G. et al. Mannose-binding protein (University of Sydney, Sydney, Australia); Bằng Đức Nguyễn,
148. Narotam, P. K. et al. Hyponatremic natriuretic B allele confers protection against tuberculous Yến Bích Nguyễn (Pham Ngoc Thach Hospital for
syndrome in tuberculous meningitis: the probable role meningitis. Pediatr. Res. 45, 459–464 (1999). Tuberculosis & Lung Diseases, Ho Chi Minh City, Vietnam);
of atrial natriuretic peptide. Neurosurgery 34, 173. Rizvi, I. et al. Vitamin D status, vitamin D receptor and Vinod Patel (University of KwaZulu-Natal, Durban, South
982–988 (1994). toll like receptor‑2 polymorphisms in tuberculous Africa); Thomas Pouplin (Mahidol-Oxford Tropical Medicine
149. Sterns, R. H. & Silver, S. M. Cerebral salt wasting meningitis: a case-control study. Infection 44, 633–640 Research Unit, Bangkok, Thailand); Lalita Ramakrishnan
versus SIADH: what difference? J. Am. Soc. Nephrol. (2016). (University of Cambridge, Cambridge, UK); Johan F.
19, 194–196 (2008). 174. Zhao, Y. et al. Genetic polymorphisms of CCL1 Schoeman, Regan Solomons, Ronald Van Toorn (University
150. Moller, K., Larsen, F. S., Bie, P. & Skinhoj, P. The rs2072069 G/A and TLR2 rs3804099 T/C in of Stellenbosch, Cape Town, South Africa); James Seddon
syndrome of inappropriate secretion of antidiuretic pulmonary or meningeal tuberculosis patients. Int. (Imperial College, London, UK); Javeed Shah (University of
hormone and fluid restriction in meningitis—how J. Clin. Exp. Pathol. 8, 12608–12620 (2015). Washington, Washington, USA); Jaya S. Tyagi (All India

NATURE REVIEWS | NEUROLOGY ADVANCE ONLINE PUBLICATION | 17

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Institute of Medical Sciences, New Delhi, India); Douwe H. Foundation Of South Africa (96841). G.T. is supported by R.J.W., U.R., U.K.M., R.v.C., wrote the article, and all authors
Visser (VU University Medical Center, Amsterdam, The the Wellcome Trust through a Major Overseas Programme contributed to the review and editing of the manuscript
Netherlands); Robert J. Wilkinson (Imperial College and grant (106680/Z/14/Z) and an Investigator Award before submission.
The Francis Crick Institute, London, UK and University of (110179/Z/15/Z).
Cape Town, South Africa). R.J.W. is supported by the Francis Competing interests statement
Crick Institute, which receives its core funding from Cancer Author contributions The authors declare no competing interests.
Research UK (FC00110218), the UK Medical Research R.J.W, U.R., U.K.M., R.v.C., N.T.H.M., K.E.D., M.C., A.F., and
C o u n c i l ( F C 0 0110 21 8 ) , a n d t h e We l l c o m e Tr u s t G.T. researched data for the article, R.J.W, U.R., U.K.M., Publisher’s note
(FC00110218). He also receives support from the Wellcome R.v.C., K.E.D., M.C., A.F., R. Savic, R. Solomons, and G.T Springer Nature remains neutral with regard to jurisdictional
Trust (104803, 203135) and the National Research made a substantial contribution to discussion of content, claims in published maps and institutional affiliations.

18 | ADVANCE ONLINE PUBLICATION www.nature.com/nrneurol

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.