Series

Headache 2
The pathophysiology of migraine: implications for clinical
management
Andrew Charles

The understanding of migraine pathophysiology is advancing rapidly. Improved characterisation and diagnosis of its Lancet Neurol 2017
clinical features have led to the view of migraine as a complex, variable disorder of nervous system function rather Published Online
than simply a vascular headache. Recent studies have provided important new insights into its genetic causes, December 8, 2017
http://dx.doi.org/10.1016/
anatomical and physiological features, and pharmacological mechanisms. The identification of new migraine-
S1474-4422(17)30435-0
associated genes, the visualisation of brain regions that are activated at the earliest stages of a migraine attack, a
UCLA Goldberg Migraine
greater appreciation of the potential role of the cervical nerves, and the recognition of the crucial role for neuropeptides Program, Department of
are among the advances that have led to novel targets for migraine therapy. Future management of migraine will have Neurology, David Geffen
the capacity to tailor treatments based on the distinct mechanisms of migraine that affect individual patients. School of Medicine at UCLA,
Los Angeles, CA, USA
(Prof A Charles MD)
Introduction on only 8 of these 15 headache days, and that medication
Correspondence to:
The ongoing Global Burden of Diseases, Injuries, and overuse can exist concurrently with a diagnosis of Prof Andrew Charles, UCLA
Risk Factors Study continues to identify migraine as a chronic migraine. Goldberg Migraine Program,
leading cause of disability worldwide,1,2 particularly in This parsing of headache days with or without migraine Department of Neurology,
individuals younger than 50 years.3 Notably, disorders features illustrates how migraine diagnosis can be David Geffen School of Medicine
at UCLA, Los Angeles,
that are commonly comorbid with migraine, including confounded by the variability of attack features from CA 90095, USA
neck pain, depression, and anxiety, are also among the person to person and from attack to attack in the same [email protected]
top ten causes of disability worldwide, placing migraine individual.5,6 Patients with migraine commonly report that
in a central position among the world’s most disabling they have more than one type of headache, with each type
disorders.1,2 Furthermore, medication overuse headache having considerably different clinical features. For
is now one of the top causes of disability worldwide.1 The example, prospective recording of symptoms indicates
overlap between migraine and several other commonly that the occurrence of premonitory symptoms, nausea,
disabling disorders indicates that they might have and aura can be highly variable in an individual.5,6
common mechanisms; improved understanding of Medication overuse can introduce additional variability,
these shared mechanisms could inform the clinical and approximately 50% of patients with chronic migraine
management of the diseases that cause a substantial revert to episodic migraine after drug withdrawal.7
proportion of the world’s leading disability. Individual migraine attacks with different clinical
Much progress has been made in the understanding of characteristics might respond differently to acute or
migraine, which has created new opportunities for more preventive therapies;8 thus, recording such variability in
effective management of patients. In this Series paper, clinical features and medication use could be particularly
I describe the most recent advances in genetics and important in analysis of clinical trial data in which nausea
pathophysiology, and their implications in the or aura symptoms are the defining features of migraine
management of patients with migraine. attacks, as well as in epidemiological studies in which
patients are classified as having only one diagnosis (such
Classification and diagnosis as migraine with aura vs without aura, or chronic migraine
The International Classification of Headache Disorders vs medication overuse headache). The development of
(ICHD) remains an invaluable resource for criteria genetic, biochemical, and imaging biomarkers could, in
for the diagnosis of migraine. The evolution of this combination with detailed characterisation of clinical
classification system reflects a growing understanding features, lead to a more accurate diagnosis of migraine
of the heterogeneity of headache disorders and their and a better ability to predict an individual patient’s
variable clinical presentations. Several important response to different therapies.
revisions regarding migraine diagnosis have been Another important change in the ICHD-3 beta criteria
made in the most recent iteration of the ICHD (ICHD-3 is their updated approach to diagnosing vestibular
beta4). For example, using the ICHD-2 criteria, patients symptoms, including vertigo and dizziness, which are
needed to have 15 migraine headache days per month common features of a migraine attack. These symptoms
in the absence of medication overuse to receive a were previously included as part of the diagnosis of
diagnosis of chronic migraine. ICHD-3 beta now basilar migraine, which has been replaced by the
specifies that patients must have 15 headache days per diagnosis of migraine with brainstem aura.4 The
month, but migraine-associated features are required different nomenclature is in part an acknowledgment of

www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0 1

 Series

Factors
the appendix of ICHD-3 beta, which describes new
disorders that need further validation.10 Depending on
Genes Environment Metabolism
>38 migraine- Barometric pressure Diet other associated symptoms, migraine-associated vertigo
associated gene Stress Neuroendocrine can now be classified either as vestibular migraine or
polymorphisms function
migraine with brainstem aura. Regardless of the
classification, migraine is commonly associated with
Hormones Drugs
Menstrual cycle Exacerbating episodic vertigo.10 An increased understanding of how
Pregnancy medications migraine leads to derangements of vestibular function
will enable better diagnosis and management of patients
Mechanisms
with episodic vertigo.
Hypothalamic activation

Alteration in
Genetics
thalamo-cortical circuits Although migraine is commonly familial, genome-wide
association studies have not yet identified any genetic
Altered brain connectivity
alterations with large effect sizes.11 Single gene mutations
have been found in rare migraine syndromes,
Brainstem activation
including familial hemiplegic migraine and monogenic
vasculopathies,4,12–14 as well as in individual families. Two
Cortical spreading genetic mutations associated with migraine identified in
depolarisation
two families are in the gene encoding the enzyme casein
Release of kinase 1δ (CK1δ). These mutations are also associated
CGRP and with familial advanced sleep phase syndrome, a circadian
PACAP
rhythm disorder in which sleep onset and awakening
from sleep are shifted to earlier hours.15 This disorder is
Migraine attack
Migraine genes
considered to be rare; however, it could be under-reported
Hormonal and metabolic state as a migraine-associated disorder because patients might
Cervical nerve anatomy adapt well to the symptoms and therefore not seek
Drugs
medical attention.
38 genomic loci associated with migraine have been
Variable attack symptoms and validated in population studies.16 The large number and
severity: premonitory, aura,
headache, and postdrome phases
functional diversity of these genes underscore the
complexity of the genetic contribution to migraine and
Figure 1: Contributing factors and mechanisms of a migraine attack the likelihood that, in most cases, migraine involves
A wide range of factors can contribute to the initiation of an attack, with interaction between multiple genes and epigenetic factors
variable mechanisms leading to a migraine attack. The clinical features of a (figure 1).11 Loci associated with migraine were found to
migraine attack then diverge on the basis of genetic, anatomical, and other be enriched in genes that are expressed in vascular and
factors. CGRP=calcitonin gene-related peptide. PACAP=pituitary adenylate
cyclase-activating polypeptide. gastrointestinal tissue, and several gene sets associated
with vascular biology (including wound healing and
the understanding that these symptoms of migraine are cell–cell interaction) were significantly enriched for loci
unlikely to be due primarily to changes in blood flow containing migraine-associated genes.16 These findings
through the basilar artery,9 but rather involve complex emphasise that although vasodilation is not the cause
changes in neural activity in the brainstem and of migraine headache,9 vascular mechanisms might
vestibular system.4 This change in classification raises nonetheless have an important role in the pathophysiology
questions about the contraindication for the use of of migraine. Another population study17 found that
triptans in basilar migraine, which was based on the genetically mediated hypercalcaemia over the lifetime is
probably incorrect presumption that these symptoms associated with an increased risk of migraine. The
were caused by constriction of the basilar artery that complex range of potential genetic mechanisms that lead
could be compounded by the vasoconstrictive action of to migraine raises the possibility that therapy could
triptans.4 Indeed, a magnetic resonance angiography eventually be tailored to specific genetic mechanisms
study9 found that administration of sumatriptan to responsible for migraine in individual patients.
19 patients during a migraine attack resulted in only
2% constriction of the basilar artery on average, a Pathophysiology
clinically insignificant vasoconstriction. The effort to Phases of a migraine attack
more accurately characterise the occurrence of vertigo A migraine attack can be divided into phases on the basis
and dizziness as symptoms of migraine attacks has also of its temporal relationship to headache: the premonitory
involved refinement of the diagnosis of vestibular phase (precedes headache), the aura phase (immediately
migraine; vestibular migraine is currently included in precedes or accompanies headache), the headache phase,

2 www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0

 Series

and the postdrome phase (after resolution of headache).18

Although this description of a migraine attack is practical, Cortex Release of CGRP and PACAP
the phases of an attack can be overlapping and variable. Cortical spreading depolarisation, altered connectivity Multiple potential sources or sites of action
Migraine aura and cognitive symptoms Headache and other symptoms
Some symptoms of a migraine attack (sensory sensitivity
Target for neuromodulation Target for small-molecule antagonists and antibodies
and neck pain) might be present throughout an attack,
whereas others (aura symptoms) might come and go.
The different phases of a migraine attack represent
an opportunity to characterise and distinguish the
physiological changes that are occurring at the start of a
migraine attack, those that are responsible for headache,
and those involved in the process of recovery (figures 1, 2).
In the premonitory phase, a variety of symptoms,
including yawning, polyuria, mood change, irritability,
light sensitivity, neck pain, and concentration difficulties,
among others,19–22 commonly occur hours before onset of
headache during a migraine attack. Although some of Thalamus
these symptoms are subjective, others, particularly sensory Sensitisation and alteration of
sensitivity, can be objectively quantified. For example, thalamo-cortical circuits
Sensory sensitivity and allodynia
changes in quantitative sensory thresholds occur hours
Target for neuromodulation
before headache, consistent with the occurrence of
Trigemino-cervical complex
subjective sensory symptoms in the premonitory phase.23,24
Pain transmission or sensitisation
PET25 and functional MRI26 studies of triggered and Hypothalamus
Headache and neck pain
Activation in premonitory phase
spontaneous migraine attacks show changes in the activity Target for medications and
Premonitory symptoms
and connectivity of the hypothalamus in the hours Target for hypothalamic peptides neuromodulation
preceding headache. These changes in hypothalamic and modulators
function could be responsible for polyuria, mood change, C1
and change in appetite preceding headache. PET studies of
Upper cervical nerves C2
the premonitory phase also show that increased activity in
Pain transmission or sensitisation C3
the occipital cortex is correlated with light sensitivity,27 and Neck pain and head pain C4
that activation of the brainstem is correlated with nausea.28 Target for local injections and C5
neuromodulation C6
Thalamic and thalamo-cortical circuits
Electrophysiological studies show changes in brain Figure 2: Anatomical sites of migraine mechanisms, symptoms, and therapeutic targets
function in the premonitory phase, particularly in the Migraine involves the simultaneous alteration in function of multiple components of the CNS and peripheral
circuits that connect the thalamus and the cortex.29 nervous system, some of which are represented in this diagram. Each of these components could be responsible
for different symptoms of migraine, and each could represent a specific therapeutic target in individual patients.
Furthermore, structural and functional imaging studies Red arrows indicate sensory inputs from the trigeminal nerve and upper cervical nerve roots, which converge in the
show differences in thalamic and thalamo-cortical activity trigemino-cervical complex. CGRP=calcitonin gene-related peptide. PACAP=pituitary adenylate cyclase-activating
in patients with migraine versus control groups both polypeptide.
during and between migraine attacks,30–32 and studies
have implicated the thalamus as an important mediator circuits, and circuits involved in anxiety and mood.
of cutaneous allodynia33 and exacerbation of headache by Although the consequences of these changes in
light.34 Taken together, these studies provide strong connectivity remain uncertain, they might be involved in
evidence that changes in thalamic and thalamo-cortical the modulation of pain and sensory sensitivity that
activity play a key part in the aberrant sensory processing occurs both ictally and interictally in patients with
that is a central feature of a migraine attack, and could migraine. Cognitive dysfunction is another common
represent a therapeutic target for pharmacological and symptom of migraine and a cause of disability associated
neuromodulatory approaches such as transcranial with migraine attacks42,43 that could be related to
magnetic stimulation.35 disruptions of normal brain functional connectivity. A
coherent pattern of changes in brain connectivity in
Network connectivity migraine has yet to emerge; however, these resting-state
Several studies have used resting-state MRI to investigate studies confirm that migraine involves widespread
changes in the connectivity of different brain regions alterations in brain function.
before and during migraine attacks. These studies
showed altered connectivity of the cortex,32,36–40 Neck pain
thalamus,30,32 hypothalamus,26 brainstem,26,37 amygdala,36,41 Neck pain is a common symptom of migraine that can
and cerebellum,38 consistent with changes in the function begin in the premonitory phase and continue through
of multiple overlapping sensory and pain-processing the postdrome phase, and might be an important

www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0 3

 Series

contributor to migraine-related disability.19,44,45 Causative the aura was apparently progressing but no visual change
structural pathology of the cervical spine is rare,46 but was perceived. These observations indicate that the
the frequent occurrence of neck pain could indicate a physiological phenomenon underlying the aura might
role for the upper cervical nerves in the transmission of have multiple foci of onset within the visual cortex, that
migraine pain. Pain inputs from the cervical nerves the clinical features of the visual aura correspond to the
converge with those from the trigeminal nerve on the specific regions of the visual cortex through which
second-order neurons in the brainstem and upper this phenomenon is propagating, and that the aura
cervical spinal cord.47 In individuals with or without phenomenon might be clinically silent in certain regions
migraine, stimulation of the cervical nerves triggers of the cortex. In rodents, which have lissencephalic
head pain, whereas C1 stimulation in patients with brains that lack the sulci and gyri of the human brain,
migraine triggers pain in a peri-orbital distribution.48 cortical spreading depolarisation propagates as a
This referral pattern could be due to central sensitisation concentric wave that commonly involves the majority of
of the trigemino-cervical complex, where cervical and one hemisphere.62 In representations of human brains, it
trigeminal inputs converge. Another contributing factor is often similarly depicted as a broad, concentric wave
could be anatomical variation of the upper cervical that traverses multiple sulci and gyri. However, mapping
nerve roots. Anatomical studies indicate considerable of the percept of the human migraine visual aura onto
variability of the structure and anastomoses of the upper models of the human visual cortex indicates that the
cervical nerve roots, particularly the C1 root in human physiological phenomenon that causes the migraine
beings.49 This variability raises the intriguing possibility aura travels in a much more spatially limited and linear
that structural differences in the cervical nerve roots manner along single gyri and sulci,62 similar to the
could influence the pattern of migraine pain (including pattern that is observed in recordings of spreading
headache), and particularly its response to local depolarisations in patients with brain ischaemia or
therapies such as suboccipital injections of local traumatic injury.63
anaesthetics and steroids. A nerve-tracing study50 A long-standing hypothesis is that the aura is a primary
showed that branches of the trigeminal nerve can reach initiator of a migraine attack, leading to headache and
the neck musculature through the skull, suggesting a other migraine symptoms. This hypothesis is based in
possible role for trigeminal afferents in migraine-related part on work in animal models, in which cortical
neck pain.50 spreading depolarisation (which was previously
described as cortical spreading depression), believed to
Aura be the physiological substrate of the migraine aura, was
The clinical significance of migraine aura and its shown to activate pain signalling via both peripheral
underlying mechanisms continue to be topics of active trigeminal and central descending pathways.62 Whether
investigation. Population studies indicate that a diagnosis cortical spreading depolarisation and migraine aura
of migraine with aura is associated with an increased risk cause headache in human beings, however, remains
of other comorbidities, such as patent foramen ovale,51 controversial. Migraine aura commonly occurs without
ischaemic stroke including perioperative stroke,52,53 headache, and most migraine attacks do not include
restless legs syndrome,54,55 Parkinson’s disease,56,57 bipolar aura, indicating that the aura is not necessary or
disorder,58 and panic disorder.59 As with other migraine sufficient for headache. Prospective patient reporting of
symptoms, the occurrence of an aura during a migraine migraine attack features also reveals that a majority of
attack is variable for most patients, and the clinical patients might have headache as well as other defining
features of the aura itself might also vary considerably in migraine symptoms, including light sensitivity and
a patient.5,6 Although visual symptoms are prevalent, nausea, at the time of aura onset,64 raising questions
sensory, language, and olfactory symptoms are also about the role of aura mechanisms as initiators of
common,5,6 and might occur either in conjunction with these symptoms. An alternative hypothesis is that
or independently of visual symptoms. The classic cortical spreading depolarisation represents just one
scintillating scotoma of migraine occurs in approximately component—not necessarily the primary component—
50% of patients; flashing lights, scotoma without of the widespread and variable dysfunction of the
scintillation, or non-descript distortion of vision are also nervous system that comprises the aberrant brain state
commonly reported.5,60 of a migraine attack.65 Cortical spreading depolarisation
Systematic recording of visual aura by a single in rodents is associated with brain tissue swelling and
individual over nearly two decades has provided has been reported to cause closure of perivascular
important insights into the initiation and propagation of spaces, a phenomenon that is hypothesised to impair
the aura phenomenon in the visual cortex.61 In this the physiological clearance of solutes.66 The relevance of
individual, the location of onset of the aura within the this phenomenon to migraine remains uncertain given
visual field was variable, the characteristics of the visual the major differences between cortical spreading
percept changed from a positive wave front to a negative depolarisation in the lissencephalic rodent cortex and in
scotoma as the aura progressed, and during certain times the gyrencephalic human cortex, as well as the fact that

4 www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0

 Series

migraine aura can occur repetitively and frequently over The evidence for pituitary adenylate cyclase-activating
a lifetime with no clear deleterious effects on brain polypeptide (PACAP) as a mediator of migraine parallels
structure or function.61 that for CGRP. Like CGRP, systemic administration of
PACAP triggers migraine in susceptible individuals,76
Neuropeptides as mediators of migraine and elevated concentrations of PACAP have been
Accumulating evidence indicates a primary role for reported in patients with migraine during attacks.77
calcitonin gene-related peptide (CGRP) as a mediator of Generalised flushing and sustained vasodilation are
migraine and as an important therapeutic target. Studies common responses to administration of PACAP,
have shown that CGRP is released into the circulation whereas these responses do not occur as prominently
during a migraine or cluster headache attack, and that its with CGRP.76 Therefore, PACAP, like CGRP, represents
concentration normalises with triptan therapy but not with a promising therapeutic target in migraine, and future
a non-specific opioid analgesic.67 CGRP concentrations therapeutic approaches targeting PACAP will validate
have been reported to be persistently elevated in patients whether it is also an important mediator of migraine.
with chronic migraine.68 Infusion of CGRP triggers
delayed migraine in susceptible individuals,69 and several Management
small-molecule CGRP antagonists have shown efficacy as Effective clinical management of migraine requires
acute migraine therapies.70,71 recognition and elimination of specific exacerbating
Notably, an early study67 of peptide release in 22 patients factors, and personalisation of acute and preventive
with migraine showed that the concentrations of treatment approaches.
vasoactive intestinal peptide, substance P, and
neuropeptide Y were not elevated during migraine Lifestyle factors
attacks. This observation means that CGRP release is not Consistency of diet, an adequate amount of sleep,
a component of generalised neurogenic inflammation, consistent caffeine intake, and regular exercise are all
which, as previously defined based on animal models, is approaches to reduce migraine frequency and severity,
primarily mediated by substance P.72 Furthermore, although no high-quality research has supported the
several substance P receptor antagonists that are potent efficacy of these approaches yet. Patients with migraine
inhibitors of neurogenic inflammation in animal models often focus on identifying dietary triggers for their
showed no efficacy as migraine therapies in clinical attacks, but no high-quality evidence exists to support
trials.73 Therefore, if the term neurogenic inflammation the efficacy of any specific elimination diet.78 Instead,
is to be used to describe mechanisms underlying patient-identified triggers could be a reflection of
migraine in human beings, the fact that it is clearly a symptoms that are occurring at the earliest stages of a
different phenomenon from what is observed in animal migraine attack; sensitivity to light, sound, or smell
models needs to be taken into account. might precede headache during a migraine attack,
The reported efficacy of antibodies targeting CGRP or leading patients to the conclusion that bright light, loud
its receptors is important for the understanding of the sounds, or strong smells are triggers rather than
pathophysiology of migraine. First, because CGRP is symptoms of the premonitory phase.20
unlikely to cross the blood–brain barrier in substantial
concentrations, targeting of CGRP outside the brain Exacerbating medications
might prevent migraine. Although antibodies can target Various medications might exacerbate migraine. The
CGRP or its receptors at brain regions that might be frequent use of acute migraine medications can be
outside the blood–brain barrier, such as the median associated with worsening of migraine, and the ICHD-3
eminence, area postrema, and pineal gland,74 their beta4 has established different numerical thresholds for
therapeutic action might be entirely peripheral (including the diagnosis of different types of medication overuse
the trigeminal ganglion, which is outside the blood–brain headache. Medication overuse is classified as at least
barrier). The CGRP antagonist fremanezumab inhibits 10 days per month using ergotamines, triptans,
activation of central trigeminovascular neurons with combination analgesics, or opioids for at least 3 months,
input from the intracranial dura, but not the facial skin and at least 15 days per month using simple analgesics
or cornea,75 providing evidence that antibodies against (eg, non-steroidal anti-inflammatory drugs and
CGRP can inhibit trigeminal neurons. However, their paracetamol) for at least 3 months. This designation is
site of action along the trigeminal pathway remains controversial, particularly given that the so-called
uncertain. Identification of the site (or sites) of action tipping point for medications as a risk factor for
of the antibodies is an important goal for improved transition from episodic to chronic migraine is as few
understanding of the basic mechanisms of migraine and as 5 days per month for barbiturates (butalbital
for development of new therapies. Regardless of their complex) or 8 days per month for opioids.79 Indeed,
site of action, the efficacy of monoclonal antibodies each of the acute medications potentially causing
against CGRP or its receptors for migraine prevention medication overuse appears to have a different profile
confirms the primary role of CGRP in migraine. in terms of its risk of causing progression of migraine.79

www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0 5

 Series

Further studies are needed to generate evidence-based department.86 The substance P antagonist aprepitant has
guidelines for appropriate limits for the use of acute been reported to be effective in the therapy of nausea
migraine medications. associated with administration of dihydroergotamine87
Medications taken for other indications might also and for cyclic vomiting syndrome in children.88 Thus,
worsen migraine. Oral contraceptive preparations, substance P antagonists might be effective in the
post-menopausal hormone replacement, decongestants, treatment of migraine-associated nausea.
selective serotonin reuptake inhibitors, and proton-pump
inhibitors are among the commonly prescribed Migraine aura
medications that are observed anecdotally in clinical Compared with migraine without aura, attacks with aura
practice to worsen migraine. The scarcity of data on this might be less responsive to triptans and might also have a
topic is surprising, given how common the exacerbation differential response to preventive therapies.8 Currently,
of migraine by other medications can be. Prospective no evidence-based treatment for migraine with aura is
studies to obtain such data are unlikely to be done, but available, although results from a clinical trial of ketamine
high-quality population data from registries could have for prolonged aura suggest that glutamate receptor
the potential to delineate the adverse effects of different antagonists might be effective.89 Migraine with aura
medications on migraine frequency and severity. seems to be related to stroke, potentially through the
association between migraine aura and right-to-left shunt
Acute treatment in the systemic circulation, including that caused by
The overall pharmacological approach to the management patent foramen ovale.51,90 Thus, medications that cause
of acute migraine has not changed much since 2012. hypercoagulability (eg, oral contraceptives) could increase
Triptans, non-steroidal anti-inflammatory drugs, and the risk of stroke due to an increased risk of paradoxical
antiemetics continue to be the mainstays of acute embolism.91 However, most data regarding stroke risk
migraine therapy.80,81 Different non-oral and oral with oral contraceptives were obtained when oestrogen
preparations of triptans and non-steroidal anti- doses in these preparations were much higher than they
inflammatory drugs have been developed (nasal sprays, are currently. Similarly, no evidence exists indicating that
different injection preparations, transdermal patches, triptan use should be contraindicated in patients with
and oral powder formulations). Efficacy might be migraine with aura.
improved because of the more rapid onset of action than
that of existing medications and, in the case of non-oral Preventive treatment
preparations, because they avoid the issue of impaired As with acute therapies, little overall change has occurred
gastrointestinal absorption during migraine attacks.82 in migraine preventive therapy since 2012. Beta blockers,
Administration of acute migraine therapy as early as tricyclic antidepressants, anticonvulsants including
possible within a migraine attack remains an important topiramate and divalproex sodium, onabotulinum toxin
principle for optimisation of efficacy. An interesting A (for chronic migraine), and flunarizine (outside the
question that arises from the increased recognition—by USA) continue to be standard therapies for migraine
both patients and researchers—of premonitory symptoms prevention.92,93 A randomised placebo-controlled study94
is whether acute treatment during the premonitory phase of candesartan as an effective preventive therapy in
could result in even greater efficacy for some patients. 72  patients with migraine has led to increasing use of
this treatment, particularly because of its generally
Nausea excellent tolerability. Unfortunately, clinical criteria or
Nausea is a highly disabling symptom on its own, and biomarkers that predict which specific preventive therapy
therefore the relationship between migraine-associated might be most effective for an individual patient are not
nausea and greater disability is not surprising.83 Patients available, and the choice of preventive therapy is therefore
commonly have nausea with some, but not all, of their often based on tolerability or comorbid conditions.
attacks,5,6 and migraine-associated nausea is generally Injections of local anaesthetics with or without steroids,
undertreated. Severe nausea is associated with reduced particularly in the region of occipital nerves, are routinely
therapeutic efficacy of intravenous dihydroergotamine in done in headache centres as preventive therapy for
the inpatient setting,84 and a population-based study85 patients with migraine. These procedures are often
indicated that persistent frequent nausea is a risk factor described as nerve blocks, although blockade of nerve
for progression of migraine, suggesting that effective sensory function as evidenced by anaesthesia in the
treatment of nausea could improve both short-term and distribution of the nerve might not be required for
long-term outcomes of migraine therapy. Accumulating therapeutic efficacy. Clinical experience indicates that
evidence supports the efficacy of dopamine receptor these procedures are particularly beneficial for patients
antagonists, which are generally thought of as antiemetic with very frequent migraine or status migrainosis.
therapies, as acute migraine therapies. Phenothiazine and An observational study95 and clinical experience have
metoclopramide antiemetic therapies were efficacious as indicated that tenderness over the occipital nerve and
treatment for migraine in the setting of the emergency forward radiation of pain on pressure over the occipital

6 www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0

 Series

Declaration of interests
Search strategy and selection criteria I have served on advisory boards for Alder BioPharmaceuticals,
Biohaven, eNeura, and Eli Lilly, as a compensated consultant for Amgen,
References were identified by searches of PubMed between as a symposium speaker for Amgen and Eli Lilly, and received grant
Jan 1, 2012, and Aug 31, 2017, and references from relevant support from Takeda.
articles. The search terms “migraine”, “diagnosis”, “disability”, References
“genetics”, “pathophysiology”, “PET”, “MRI”, 1 GBD 2015 Disease and Injury Incidence and Prevalence
Collaborators. Global, regional, and national incidence, prevalence,
“electrophysiology”, “pharmacology”, “premonitory”, “aura”, and years lived with disability for 310 diseases and injuries,
“postdrome”, “therapy”, “treatment”, “CGRP”, and “PACAP” 1990–2015: a systematic analysis for the Global Burden of Disease
were used. I did not apply any language restrictions. The final Study 2015. Lancet 2016; 388: 1545–602.
2 GBD 2016 Disease and Injury Incidence and Prevalence
reference list was generated on the basis of relevance to the Collaborators. Global, regional, and national incidence, prevalence,
topics covered in this Series paper. and years lived with disability for 328 diseases and injuries for
195 countries, 1990–2016: a systematic analysis for the Global
Burden of Disease Study 2016. Lancet 2017; 390: 1211–59.
3 Steiner TJ, Stovner LJ, Vos T. GBD 2015: migraine is the third cause
nerve are predictors of benefit from occipital injections. of disability in under 50s. J Headache Pain 2016; 17: 104.
Results from randomised placebo-controlled studies 4 Headache Classification Committee of the International Headache
have been mixed: one study96 of 69 patients with episodic Society (IHS). The International Classification of Headache
Disorders, 3rd edition (beta version). Cephalalgia 2013; 33: 629–808.
or chronic migraine showed no benefit, whereas
5 Hansen JM, Goadsby PJ, Charles AC. Variability of clinical features
two studies97,98 of 84 and 36 patients with chronic migraine in attacks of migraine with aura. Cephalalgia 2016; 36: 216–24.
did show benefit. Positive responses to occipital nerve 6 Viana M, Sances G, Ghiotto N, et al. Variability of the characteristics
injections were also reported in patients with prolonged of a migraine attack within patients. Cephalalgia 2016; 36: 825–30.
7 Munksgaard SB, Bendtsen L, Jensen RH. Detoxification of
or persistent migraine aura,99 as well as in those with medication-overuse headache by a multidisciplinary treatment
triptan overuse headache.100 Because the occipital nerve is programme is highly effective: a comparison of two consecutive
believed to be composed primarily of C2 and C3 afferents, treatment methods in an open-label design. Cephalalgia 2012;
32: 834–44.
occipital injection could therefore miss the C1 nerve,
8 Hansen JM, Goadsby PJ, Charles A. Reduced efficacy of
which is also a potential modulator of migraine pain.48 sumatriptan in migraine with aura vs without aura. Neurology 2015;
Occipital nerve injections are relatively easy to do, with 84: 1880–85.
low complication rates and, thus, are a reasonable 9 Amin FM, Asghar MS, Hougaard A, et al. Magnetic resonance
angiography of intracranial and extracranial arteries in patients with
approach to try in selected patients, despite mixed spontaneous migraine without aura: a cross-sectional study.
supporting evidence. Clinical trials with refinement of Lancet Neurol 2013; 12: 454–61.
predictive clinical criteria would be helpful in guiding 10 von Brevern M, Lempert T. Vestibular migraine. Handb Clin Neurol
2016; 137: 301–16.
appropriate use of occipital injections in the future. 11 Nyholt DR, van den Maagdenberg AM. Genome-wide association
studies in migraine: current state and route to follow.
Conclusions and future directions Curr Opin Neurol 2016; 29: 302–08.
12 Sutherland HG, Griffiths LR. Genetics of migraine: insights into
The pace of progress in the understanding of migraine is the molecular basis of migraine disorders. Headache 2017;
accelerating rapidly, and the direct translation of the 57: 537–69.
results of basic migraine research into new treatments 13 Tan RY, Markus HS. CADASIL: migraine, encephalopathy, stroke
that specifically target migraine mechanisms is and their inter-relationships. PLoS One 2016; 11: e0157613.
14 Stam AH, Kothari PH, Shaikh A, et al. Retinal vasculopathy with
particularly gratifying. Improved understanding of cerebral leukoencephalopathy and systemic manifestations.
genetic factors underlying migraine has the potential to Brain 2016; 139: 2909–22.
tailor therapies to individuals with different genetic 15 Brennan KC, Bates EA, Shapiro RE, et al. Casein kinase Iδ
mutations in familial migraine and advanced sleep phase.
backgrounds. Characterisation of the premonitory phase Sci Transl Med 2013; 5: 183ra56.
of a migraine attack and its underlying mechanisms 16 Gormley P, Anttila V, Winsvold BS, et al. Meta-analysis of
provides an opportunity to treat a migraine attack at its 375,000 individuals identifies 38 susceptibility loci for migraine.
Nat Genet 2016; 48: 856–66.
very earliest stages, and could reveal new therapeutic
17 Yin P, Anttila V, Siewert KM, Palotie A, Davey Smith G, Voight BF.
targets, such as the hypothalamus. Neuromodulation Serum calcium and risk of migraine: a Mendelian randomization
approaches could enable physicians to specifically target study. Hum Mol Genet 2017; 26: 820–28.
novel central and peripheral migraine mechanisms, such 18 Charles A. The evolution of a migraine attack—a review of recent
evidence. Headache 2013; 53: 413–19.
as alterations in thalamo-cortical circuits, or contributions 19 Lampl C, Rudolph M, Deligianni CI, Mitsikostas DD. Neck pain in
of cervical nerve roots to migraine headache. The episodic migraine: premonitory symptom or part of the attack?
recognition of the key roles for neuropeptides, including J Headache Pain 2015; 16: 566.
CGRP and PACAP, and the development of therapies 20 Schulte LH, Jürgens TP, May A. Photo-, osmo- and phonophobia in
the premonitory phase of migraine: mistaking symptoms for
targeting these peptides or their receptors represent a triggers? J Headache Pain 2015; 16: 14.
promising new approach to migraine treatment. Now is 21 Karsan N, Prabhakar P, Goadsby PJ. Characterising the premonitory
a crucial time to maintain the momentum that has been stage of migraine in children: a clinic-based study of 100 patients in
a specialist headache service. J Headache Pain 2016; 17: 94.
established by recent basic and clinical research, and 22 Laurell K, Artto V, Bendtsen L, et al. Premonitory symptoms in
push forward towards safe, efficacious, and individualised migraine: a cross-sectional study in 2714 persons. Cephalalgia 2016;
care for patients with migraine. 36: 951–59.

www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0 7

 Series

23 Schwedt TJ, Zuniga L, Chong CD. Low heat pain thresholds in 47 Goadsby PJ, Knight YE, Hoskin KL. Stimulation of the greater
migraineurs between attacks. Cephalalgia 2015; 35: 593–99. occipital nerve increases metabolic activity in the trigeminal nucleus
24 Uglem M, Omland PM, Nilsen KB, et al. Does pain sensitivity caudalis and cervical dorsal horn of the cat. Pain 1997; 73: 23–28.
change by migraine phase? A blinded longitudinal study. 48 Johnston MM, Jordan SE, Charles AC. Pain referral patterns of
Cephalalgia 2016; published online Nov 15. DOI:333102416679955. the C1 to C3 nerves: implications for headache disorders.
25 Maniyar FH, Sprenger T, Monteith T, Schankin C, Goadsby PJ. Ann Neurol 2013; 74: 145–48.
Brain activations in the premonitory phase of 49 Kemp WJ 3rd, Tubbs RS, Cohen-Gadol AA. The innervation of the
nitroglycerin-triggered migraine attacks. Brain 2014; 137: 232–41. cranial dura mater: neurosurgical case correlates and a review of the
26 Schulte LH, May A. The migraine generator revisited: literature. World Neurosurg 2012; 78: 505–10.
continuous scanning of the migraine cycle over 30 days and 50 Schueler M, Neuhuber WL, De Col R, Messlinger K. Innervation of
three spontaneous attacks. Brain 2016; 139: 1987–93. rat and human dura mater and pericranial tissues in the
27 Maniyar FH, Sprenger T, Schankin C, Goadsby PJ. parieto-temporal region by meningeal afferents. Headache 2014;
Photic hypersensitivity in the premonitory phase of migraine—a 54: 996–1009.
positron emission tomography study. Eur J Neurol 2014; 21: 1178–83. 51 Takagi H, Umemoto T, for the ALICE (All-Literature Investigation
28 Maniyar FH, Sprenger T, Schankin C, Goadsby PJ. The origin of of Cardiovascular Evidence) Group. A meta-analysis of case-control
nausea in migraine—a PET study. J Headache Pain 2014; 15: 84. studies of the association of migraine and patent foramen ovale.
29 de Tommaso M, Ambrosini A, Brighina F, et al. Altered processing J Cardiol 2016; 67: 493–503.
of sensory stimuli in patients with migraine. Nat Rev Neurol 2014; 52 Timm FP, Houle TT, Grabitz SD, et al. Migraine and risk of
10: 144–55. perioperative ischemic stroke and hospital readmission:
30 Hodkinson DJ, Wilcox SL, Veggeberg R, et al. Increased amplitude hospital based registry study. BMJ 2017; 356: i6635.
of thalamocortical low-frequency oscillations in patients with 53 Li L, Schulz UG, Kuker W, Rothwell PM, for the Oxford Vascular
migraine. J Neurosci 2016; 36: 8026–36. Study. Age-specific association of migraine with cryptogenic TIA
31 Coppola G, Di Renzo A, Tinelli E, et al. Thalamo-cortical network and stroke: population-based study. Neurology 2015; 85: 1444–51.
activity between migraine attacks: insights from MRI-based 54 Lin GY, Lin YK, Lee JT, et al. Prevalence of restless legs syndrome in
microstructural and functional resting-state network correlation migraine patients with and without aura: a cross-sectional,
analysis. J Headache Pain 2016; 17: 100. case-controlled study. J Headache Pain 2016; 17: 97.
32 Coppola G, Di Renzo A, Tinelli E, et al. Thalamo-cortical network 55 Schürks M, Winter A, Berger K, Kurth T. Migraine and restless legs
activity during spontaneous migraine attacks. Neurology 2016; syndrome: a systematic review. Cephalalgia 2014; 34: 777–94.
87: 2154–60. 56 Wang HI, Ho YC, Huang YP, Pan SL. Migraine is related to an
33 Wang T, Chen N, Zhan W, et al. Altered effective connectivity of increased risk of Parkinson’s disease: a population-based,
posterior thalamus in migraine with cutaneous allodynia: propensity score-matched, longitudinal follow-up study.
a resting-state fMRI study with Granger causality analysis. Cephalalgia 2016; 36: 1316–23.
J Headache Pain 2015; 17: 17. 57 Scher AI, Ross GW, Sigurdsson S, et al. Midlife migraine and late-life
34 Noseda R, Bernstein CA, Nir RR, et al. Migraine photophobia parkinsonism: AGES-Reykjavik study. Neurology 2014; 83: 1246–52.
originating in cone-driven retinal pathways. Brain 2016; 139: 1971–86. 58 Saunders EF, Nazir R, Kamali M, et al. Gender differences, clinical
35 Andreou AP, Holland PR, Akerman S, Summ O, Fredrick J, correlates, and longitudinal outcome of bipolar disorder with
Goadsby PJ. Transcranial magnetic stimulation and potential comorbid migraine. J Clin Psychiatry 2014; 75: 512–19.
cortical and trigeminothalamic mechanisms in migraine. 59 Smitherman TA, Kolivas ED, Bailey JR. Panic disorder and
Brain 2016; 139: 2002–14. migraine: comorbidity, mechanisms, and clinical implications.
36 Chong CD, Gaw N, Fu Y, Li J, Wu T, Schwedt TJ. Migraine Headache 2013; 53: 23–45.
classification using magnetic resonance imaging resting-state 60 Viana M, Sances G, Linde M, et al. Clinical features of migraine
functional connectivity data. Cephalalgia 2017; 37: 828–44. aura: results from a prospective diary-aided study. Cephalalgia 2017;
37 Hougaard A, Amin FM, Larsson HB, Rostrup E, Ashina M. 37: 979–89.
Increased intrinsic brain connectivity between pons and 61 Hansen JM, Baca SM, Vanvalkenburgh P, Charles A. Distinctive
somatosensory cortex during attacks of migraine with aura. anatomical and physiological features of migraine aura revealed by
Hum Brain Mapp 2017; 38: 2635–42. 18 years of recording. Brain 2013; 136: 3589–95.
38 Amin FM, Hougaard A, Magon S, et al. Change in brain network 62 Charles AC, Baca SM. Cortical spreading depression and migraine.
connectivity during PACAP38-induced migraine attacks: Nat Rev Neurol 2013; 9: 637–44.
a resting-state functional MRI study. Neurology 2016; 86: 180–87. 63 Hartings JA, Shuttleworth CW, Kirov SA, et al. The continuum of
39 Tedeschi G, Russo A, Conte F, et al. Increased interictal visual spreading depolarizations in acute cortical lesion development:
network connectivity in patients with migraine with aura. examining Leão’s legacy. J Cereb Blood Flow Metab 2017; 37: 1571–94.
Cephalalgia 2016; 36: 139–47. 64 Hansen JM, Lipton RB, Dodick DW, et al. Migraine headache is
40 Niddam DM, Lai KL, Fuh JL, Chuang CY, Chen WT, Wang SJ. present in the aura phase: a prospective study. Neurology 2012;
Reduced functional connectivity between salience and visual 79: 2044–49.
networks in migraine with aura. Cephalalgia 2016; 36: 53–66. 65 Charles A. Migraine: a brain state. Curr Opin Neurol 2013;
41 Chen Z, Chen X, Liu M, Dong Z, Ma L, Yu S. Altered functional 26: 235–39.
connectivity of amygdala underlying the neuromechanism of 66 Schain AJ, Melo-Carrillo A, Strassman AM, Burstein R.
migraine pathogenesis. J Headache Pain 2017; 18: 7. Cortical spreading depression closes paravascular space and
42 Santangelo G, Russo A, Trojano L, et al. Cognitive dysfunctions and impairs glymphatic flow: implications for migraine headache.
psychological symptoms in migraine without aura: a cross-sectional J Neurosci 2017; 37: 2904–15.
study. J Headache Pain 2016; 17: 76. 67 Goadsby PJ, Edvinsson L, Ekman R. Vasoactive peptide release in
43 Gil-Gouveia R, Oliveira AG, Martins IP. The impact of cognitive the extracerebral circulation of humans during migraine headache.
symptoms on migraine attack-related disability. Cephalalgia 2016; Ann Neurol 1990; 28: 183–87.
36: 422–30. 68 Cernuda-Morollón E, Larrosa D, Ramón C, Vega J,
44 Ashina S, Bendtsen L, Lyngberg AC, Lipton RB, Hajiyeva N, Martínez-Camblor P, Pascual J. Interictal increase of CGRP levels
Jensen R. Prevalence of neck pain in migraine and tension-type in peripheral blood as a biomarker for chronic migraine.
headache: a population study. Cephalalgia 2015; 35: 211–19. Neurology 2013; 81: 1191–96.
45 Florencio LL, Chaves TC, Carvalho GF, et al. Neck pain disability is 69 Ashina M, Hansen JM, Olesen J. Pearls and pitfalls in human
related to the frequency of migraine attacks: a cross-sectional study. pharmacological models of migraine: 30 years’ experience.
Headache 2014; 54: 1203–10. Cephalalgia 2013; 33: 540–53.
46 Viana M, Sances G, Terrazzino S, Sprenger T, Nappi G, Tassorelli C. 70 Marcus R, Goadsby PJ, Dodick D, Stock D, Manos G, Fischer TZ.
When cervical pain is actually migraine: an observational study in BMS-927711 for the acute treatment of migraine: a double-blind,
207 patients. Cephalalgia 2016; published online randomized, placebo controlled, dose-ranging trial.
Dec 7. DOI:1177/0333102416683917. Cephalalgia 2014; 34: 114–25.

8 www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0

 Series

71 Voss T, Lipton RB, Dodick DW, et al. A phase IIb randomized, 87 Chou DE, Tso AR, Goadsby PJ. Aprepitant for the management of
double-blind, placebo-controlled trial of ubrogepant for the acute nausea with inpatient IV dihydroergotamine. Neurology 2016;
treatment of migraine. Cephalalgia 2016; 36: 887–98. 87: 1613–16.
72 Tajti J, Szok D, Majláth Z, Tuka B, Csáti A, Vécsei L. Migraine and 88 Cristofori F, Thapar N, Saliakellis E, et al. Efficacy of the
neuropeptides. Neuropeptides 2015; 52: 19–30. neurokinin-1 receptor antagonist aprepitant in children with cyclical
73 May A, Goadsby PJ. Substance P receptor antagonists in the therapy vomiting syndrome. Aliment Pharmacol Ther 2014; 40: 309–17.
of migraine. Expert Opin Investig Drugs 2001; 10: 673–78. 89 Afridi SK, Giffin NJ, Kaube H, Goadsby PJ. A randomized
74 Eftekhari S, Gaspar RC, Roberts R, et al. Localization of CGRP controlled trial of intranasal ketamine in migraine with prolonged
receptor components and receptor binding sites in rhesus monkey aura. Neurology 2013; 80: 642–47.
brainstem: a detailed study using in situ hybridization, 90 Koppen H, Palm-Meinders IH, Mess WH, et al. Systemic right-to-left
immunofluorescence, and autoradiography. J Comp Neurol 2016; shunts, ischemic brain lesions, and persistent migraine activity.
524: 90–118. Neurology 2016; 86: 1668–75.
75 Melo-Carrillo A, Noseda R, Nir R, et al. Selective inhibition of 91 Tietjen GE, Collins SA. Hypercoagulability and migraine.
trigeminovascular neurons by fremanezumab: a humanized Headache 2017; published online Feb 9. DOI:1111/head.13044.
monoclonal anti-CGRP antibody. J Neurosci 2017; 37: 7149–63. 92 Silberstein SD, Holland S, Freitag F, et al. Evidence-based guideline
76 Amin FM, Hougaard A, Schytz HW, et al. Investigation of the update: pharmacologic treatment for episodic migraine prevention
pathophysiological mechanisms of migraine attacks induced by in adults: report of the Quality Standards Subcommittee of the
pituitary adenylate cyclase-activating polypeptide-38. Brain 2014; American Academy of Neurology and the American Headache
137: 779–94. Society. Neurology 2012; 78: 1337–45.
77 Zagami AS, Edvinsson L, Goadsby PJ. Pituitary adenylate cyclase 93 Pringsheim T, Davenport W, Mackie G, et al. Canadian Headache
activating polypeptide and migraine. Ann Clin Transl Neurol 2014; Society guideline for migraine prophylaxis. Can J Neurol Sci 2012;
1: 1036–40. 39: S1–59.
78 Orr SL. Diet and nutraceutical interventions for headache 94 Stovner LJ, Linde M, Gravdahl GB, et al. A comparative study of
management: a review of the evidence. Cephalalgia 2016; 36: 1112–33. candesartan versus propranolol for migraine prophylaxis:
79 Lipton RB, Serrano D, Nicholson RA, Buse DC, Runken MC, a randomised, triple-blind, placebo-controlled, double cross-over
Reed ML. Impact of NSAID and triptan use on developing chronic study. Cephalalgia 2014; 34: 523–32.
migraine: results from the American Migraine Prevalence and 95 Afridi SK, Shields KG, Bhola R, Goadsby PJ. Greater occipital nerve
Prevention (AMPP) study. Headache 2013; 53: 1548–63. injection in primary headache syndromes—prolonged effects from
80 Marmura MJ, Silberstein SD, Schwedt TJ. The acute treatment of a single injection. Pain 2006; 122: 126–29.
migraine in adults: the american headache society evidence 96 Dilli E, Halker R, Vargas B, et al. Occipital nerve block for the
assessment of migraine pharmacotherapies. Headache 2015; 55: 3–20. short-term preventive treatment of migraine: a randomized,
81 Worthington I, Pringsheim T, Gawel MJ, et al. Canadian Headache double-blinded, placebo-controlled study. Cephalalgia 2015;
Society Guideline: acute drug therapy for migraine headache. 35: 959–68.
Can J Neurol Sci 2013; 40: S1–80. 97 Inan LE, Inan N, Karadaş Ö, et al. Greater occipital nerve blockade
82 Diener HC, Charles A, Goadsby PJ, Holle D. New therapeutic for the treatment of chronic migraine: a randomized, multicenter,
approaches for the prevention and treatment of migraine. Lancet double-blind, and placebo-controlled study. Acta Neurol Scand 2015;
Neurol 2015; 14: 1010–22. 132: 270–77.
83 Lipton RB, Buse DC, Saiers J, Fanning KM, Serrano D, Reed ML. 98 Cuadrado ML, Aledo-Serrano Á, Navarro P, et al. Short-term effects
Frequency and burden of headache-related nausea: results from the of greater occipital nerve blocks in chronic migraine:
American Migraine Prevalence and Prevention (AMPP) study. a double-blind, randomised, placebo-controlled clinical trial.
Headache 2013; 53: 93–103. Cephalalgia 2017; 37: 864–72.
84 Eller M, Gelfand AA, Riggins NY, Shiboski S, Schankin C, 99 Cuadrado ML, Aledo-Serrano Á López-Ruiz P, et al.
Goadsby PJ. Exacerbation of headache during dihydroergotamine for Greater occipital nerve block for the acute treatment of prolonged or
chronic migraine does not alter outcome. Neurology 2016; 86: 856–59. persistent migraine aura. Cephalalgia 2017; 37: 812–18.
85 Reed ML, Fanning KM, Serrano D, Buse DC, Lipton RB. 100 Karadaş Ö, Özön AÖ, Özçelik F, Özge A. Greater occipital nerve
Persistent frequent nausea is associated with progression to chronic block in the treatment of triptan-overuse headache: a randomized
migraine: AMPP study results. Headache 2015; 55: 76–87. comparative study. Acta Neurol Scand 2017; 135: 426–33.
86 Orr SL, Aube M, Becker WJ, et al. Canadian Headache Society
systematic review and recommendations on the treatment of
migraine pain in emergency settings. Cephalalgia 2015; 35: 271–84.

www.thelancet.com/neurology Published online December 8, 2017 http://dx.doi.org/10.1016/S1474-4422(17)30435-0 9