Article Text


Unravelling of the paroxysmal dyskinesias
  1. Roberto Erro1,
  2. Kailash P Bhatia2
  1. 1 Center for Neurodegenerative Diseases (CEMAND), Department of Medicine, Surgery and Dentistry “Scuola Medica Salernitana”, Universitá di Salerno, Baronissi, Italy
  2. 2 Sobell Department For Motor Neuroscience and Movement Disorders, Institute of Neurology, University College London, London, UK
  1. Correspondence to Roberto Erro, Department of Medicine, Surgery and Dentistry “Scuola Medica Salernitana”,Universitá di Salerno, Baronissi 84081, Italy; rerro{at}


Paroxysmal dyskinesias (PxD) refer to a rare group of clinically and genetically heterogeneous disorders presenting with recurrent attacks of abnormal movements, typically dystonia, chorea or a combination thereof, without loss of consciousness. Classically, PxD have been categorised according to their triggers and duration of the attacks, but increasing evidence suggests that there is a certain degree of clinical and genetic overlap and challenges the concept that one phenotype is attributable to one single aetiology. Here we review the increasing spectrum of genetic conditions, as well as of other non-genetic disorders, that might present with PxD, provide criteria for case definition and propose a diagnostic workup to reach a definitive diagnosis, on which treatment is heavily dependent.

  • PRRT2
  • PNKD
  • PKD
  • PED
  • GLUT1.

Statistics from


Paroxysmal dyskinesias (PxD) are a group of heterogeneous syndromes that characteristically manifest with recurrent attacks of abnormal movements, typically dystonia, chorea or a combination thereof, without loss of consciousness.1

Considering the etymology and literal meaning of the words ‘paroxysmal’ and ‘dyskinesia’, one would realise that neither term is specific enough to unequivocally identify the entities that are classically referred to as PxD. The term ‘paroxysmal’ (from Greek paroxusmós—irritation, the severe fit of a disease) refers to sudden attack, recurrence or intensification of a disease.2 According to this definition, even waxing-and-waning conditions like tic syndromes or movement disorders that appear or worsen on action (action-myoclonus or action-tremor for instance) would fit into this category. The term dyskinesia (from Greek, dys+kinesis, altered movement) is also used in the medical literature with different meanings. Some medical dictionaries propose that it should be used to indicate an impairment of the ability to execute voluntary movements,3 whereas others emphasise the hyper-kinetic nature of the disorder and suggest that dyskinesias are involuntary jerky or slow writhing movements, often of a fixed pattern, including tics, myoclonus, chorea and dystonia.4

It appears clear that such definitions are too broad and do not match with the meaning that movement disorder experts give to the term ‘paroxysmal dyskinesias’. This calls for the need for a clear definition of what should be intended as PxD, to avoid further ambiguity in the literature.

Moreover, increasing knowledge regarding the possible aetiologies underlying the PxD has challenged the concept they could represent distinct disease entities (ie, that one phenotype could be only attributable to one single aetiology).5–7 Hence, following a brief overview of the historical aspects and the classification of PxD based on triggers (see below), a further in-depth section will be structured according to the different aetiologies that produce PxD. This section will cover novel genetic disorders that might encompass in their phenotype paroxysmal dystonia and/or chorea, but escape the current classification of PxD.

A final section is meant to provide a diagnostic strategy to deal with PxD. This framework emphasises that these diagnostic labels represent clinical syndromes (and not disease entities), provide criteria for case definition and endorse the approach recently used for dystonia in general8 (ie, isolated vs combined PxD, thereby discarding the former criterion requiring normal neurological examination between the attacks). On the basis of different clinical features including, but not limited to, triggers, appropriate investigations are suggested to reach a definite diagnosis, on which treatment is heavily dependent. It is not in the aims of the current paper to review the pathophysiological mechanisms underpinning this group of disorders and the interested readers are referred elsewhere.9 The search strategy is detailed in box 1.


Search strategy and selection criteria

References for this Review were identified by searches of PubMed, EMBASE and Google Scholar, between 1940 and January 2018, and references from both relevant articles and book chapters. The search terms ‘paroxysmal dyskinesias’, ‘paroxysmal dystonia’, ‘PRRT2’, ‘MR-1’, ‘SCL2A1’, ‘GLUT1’, ‘KNCMA1’, ‘ADCY5’, ‘ATP1A3’, ‘PKD’, ‘PNKD’ and ‘PED’ were used. There were no language restrictions. The final reference list was generated on the basis of relevance to the topics covered in this Review.

Historical definition and classification of paroxysmal dyskinesias

Between 1940 and 1977, three main forms of episodic movement disorders were recognised10–12 and classified based on the duration of attacks.12 Following this earlier classification,12 a subsequent proposal by Demirkiran and Jankovic discarded the duration of attacks as informative and focused on the difference between triggers.13 They recognised three subtypes, encompassing paroxysmal kinesigenic (PKD), non-kinesigenic (PNKD) and exercise-induced (PED) dyskinesias.13 The term dyskinesia was privileged over others previously adopted because the specific phenomenology of the attacks (ie, dystonia vs chorea) could only be presumed based on patients’ description. A fourth subtype was also proposed (ie, paroxysmal hypnogenic dyskinesias (PHD) characterised by attacks occurring during sleep without identifiable trigger),13 but this entity has been subsequently suggested to be a form of autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) in most cases.14

Each of these forms could be further stratified into primary and secondary disorders.13 However, the term ‘primary’ is increasingly dismissed as it carries the implication that there is absence of detectable abnormalities,15 whereas most primary PxD are in fact found to be secondary to a genetic defect. Moreover, it was suggested that patients with secondary PxD have interictal signs reflecting the underlying disorder,13 as opposed to cases with primary PxD. However, as anticipated above and discussed in detail below, many patients with ‘primary’ genetic forms of PxD do have interictal findings on examination. One example is represented by SLC2A1 (GLUT1, glucose transporter type 1) mutations, which can produce isolated PED (previously assigned the DYT18 number),16 PED with interictal spasticity (previously assigned the DYT9 number)17 as well as a number of other different phenotypes.

Disorders presenting with PxD


In 2011, Chen et al first reported PRRT2 mutations as the genetic cause of PKD in eight families with PKD,18 opening the way to the identification of PRRT2 mutation in several clinical syndromes previously associated with PKD such as the so-called infantile convulsions with choreoathetosis (ICCA) syndrome and benign familial infantile seizure syndrome.19 Currently, PRRT2 is the major gene accounting for PKD, with a frequency ranging from about 40% to over 90%, depending on case ascertainment.14 19 S1 S2 Onset of PxD is in childhood, very rarely later than 18 years of age. In patients with ICCA, PxD start after the onset of epilepsy (that develops within the first 2 years of life), usually after age 5, although some patients might exhibit epileptic seizures at a later age.9 14 19

Virtually all PRRT2 cases have a clear kinesigenic trigger, although in up to 40%–50% of cases anxiety, stress, startle or prolonged exercise can also induce attacks and very rarely (about 1%–2% of patients) there are no kinesigenic triggers.14 The episodes are very brief, usually lasting less than 1 min, feature both chorea and dystonia and tend to generalise.14 19 S1 S2 About half of the patients experience a sensory aura at the initial site of the attacks that patients can use to predict attacks.14 Often patients have hundreds of episodes per day, although the frequency of attacks usually decreases with advancing age after puberty and the syndrome can completely remit regardless of any treatments.14 19 S1 S2 Carbamazepine is the first-line treatment option, being very effective at low doses (50–600 md/day).14 19 S1-S4 A dramatic reduction in attacks is usually observed and this has been in fact suggested to be typical for patients with PRRT2 mutation as compared with similar cases of PKD and without such genetic defect.S3 S4

It is however now clear that PRRT2 mutations can induce additional phenotypes including episodic ataxia and migraine, often of the hemiplegic subtype.5 19 S5 As such, these features, whenever present in a single subject or in the family, make the presence of PRRT2 mutations likely.

Interestingly, PRRT2 mutations have been recently identified in 2 out of 11 patients (18.2%) with PHD,20 which supports the inclusion of PHD among the PxD. Moreover, PRRT2 mutations have been further associated with a phenotype reminiscent of benign paroxysmal torticollis of infancy (BPTI)21 and, in the case of biallelic mutations, with a complex phenotype including neurodevelopmental delay.22 Complex phenotypes including PKD along with developmental delays, intellectual disability and language abnormalities, minor dysmorphic facial features, and/or autism spectrum disorder should also raise the suspicion of 16p11.2 (micro)deletions.23 S6 In such rare cases, conventional genetic testing for PRRT2 mutations might be uninformative and microarray-based comparative genomic hybridisation has to be performed.S6 Table 1 summarises the clinical features of PRRT2 mutation as well as of the main genetic conditions producing PxD.

Table 1

Characteristics of mutations mainly associated with paroxysmal dyskinesias


In 2004, MR-1 mutations were for the first time discovered in two unrelated families with PNKD.24 Onset of the attacks is usually in the first decade.5 18 14 S7 There is always a dominant family history for similar attacks, no sporadic cases having been reported thus far.14 At the phenomenological level, attacks encompass both dystonic and choreatic features and are generalised in about 50% of the patients.14 24 S7 S8 Attacks can be seldom complicated by dysarthria, dysphagia, oculogyric crises, inability to move and/or pain and might also be fatal.14 S9 The duration of the attacks is variable but often lasts for a few hours.5 14 24 S7 S8 Although several non-kinesigenic triggers (ie, stress, tiredness, sustained exercise) can be present, in patients with MR-1 mutations attacks are characteristically brought on by coffee and/or alcohol intake5 14 24 S7 S8 as compared with patients with PNKD but without mutations in this gene.S7 MR-1 carriers do not have associated clinical features, with the exception of migraine in a few cases,5 and the interictal examination is always normal. Clonazepam is the first-line pharmacological option when lifestyle modifications (ie, avoiding coffee and alcohol) are not efficacious.14 Regardless of any treatment, there is a tendency for the attacks to reduce or remit in adulthood.5 14 S7 Rarely, MR-1 mutations can manifest with very brief attacks triggered by sudden movements, therefore resembling classic PKD.S10


Mutations in SLC2A1 encoding the GLUT1 have been discovered to cause a spectrum of neurological phenotypes, including PED.5 9 14–16 The latter is characterised by attacks of chorea and dystonia affecting mainly the lower limbs that are typically triggered by sustained exercise.5 9 14–16 25 There is some phenotype-genotype correlation, with splice site, non-sense, insertions or deletions leading to loss of function SLC2A1 mutations being associated with younger age at onset and a more severe clinical phenotype of GLUT1 deficiency syndrome, including epilepsy, hypotonia, spasticity, ataxia and developmental delay.14 25 This compares with missense SLC2A1 mutations, which more commonly present with PED in older patients,S11 the age at onset ranging from 1 to 50 years of age. As such, depending on the specific underlying genetic defect, PED can manifest as an isolated syndrome (about one-third of SLC2A1 cases) or be associated with other neurological disorders.25 S11 Most cases with PED are de novo, whereas only 10% have a positive family history.15 16 Although autosomal recessive transmission has been described in rare cases of GLUT1 deficiency syndrome,S12 this has not been reported in patients with PED. It is important to remark that while PED is the most common type of PxD reported in SLC2A1 cases, other non-kinesigenic triggers have also been described14 and patients can manifest other episodic neurological disorders including episodic ataxia.5 S13

PxD in the context of SLC2A1 cases have a positive but partial response to a ketogenic diet,26 which should be pursued to treat the underlying neuroglycopenia.


Mutations in KNCMA1, which encodes for a subunit of a calcium-activated potassium channel, have been reported in 2005 to cause a syndrome of PNKD and epilepsy.27 Clinically, the PxD borne resemblance with the non-kinesigenic variant and alcohol was noted as a possible (but not constant) trigger.27 Two additional patients with PNKD with KCNMA1 mutations have been subsequently reported to lack epilepsy, but they presented with neurodevelopmental delay.S14 As such, at variance from MR-1 patients, KCNMA1 mutations cause PNKD associated with either epilepsy or neurodevelopmental delay. PNKD in KNCMA1 carriers variably responds to antiepileptic drugs.27 S14


ECHS1 encodes for the short-chain enoyl-CoA hydratase protein, mutations of which have been reported as a cause of early-onset Leigh syndrome (or a Leigh-like syndrome with atypical, often milder form with later onset).S15 ECHS1 mutations have been further associated with PxD, which can be either isolated or combined with a number of features suggestive of a mitochondrial disease.28–30 S15 Thus, ECHS1 mutations have been associated with intermittent episodes of long-duration (30–50 min) opisthotonus with no identifiable trigger,S15 but also with episodes of dystonia clearly induced by sustained exercise.29 30 Although a previous report has labelled the latter episodes as ‘kinesigenic’,28 a careful analysis of the original case description reveals that the attacks were actually triggered by ‘physical strain’.28 Two more recent reports have confirmed that PxD due to ECHS1 mutations are more likely to be in the form of PED, with or without normal interictal neurological examination.29 30

It is worthy of note that all patients described so far have pallidal hyperintensity on T2-MRI sequences (despite being very mild in one patient with isolated PED,29 suggesting this might be a clue to suspect ECHS1 mutations. A possible benefit with ketogenic diet or a mitochondrial cocktail including thiamine, riboflavin, carnitine, coenzyme Q-10, vitamin B6 and vitamin C has been reported in PxD.28 30

Pyruvate dehydrogenase deficiency: PDHA1, PDHX and DLAT

The mitochondrial pyruvate dehydrogenase complex (PDC) catalyses the rate-limiting step in the aerobic glucose oxidation and comprises multiple copies of three subunits: pyruvate dehydrogenase (E1, encoded by the PDHA1 gene), dihydrolipoamide transacetylase (E2, encoded by the DLAT gene) and dihydrolipoamide dehydrogenase (E3), as well as an E3-binding protein (also known as component X and encoded by the PDHX gene).S16 Deficits in either subunit have been reported to cause PxD usually, but not always, embedded in complex neurologic pictures.31 32 S17 S18

Mutations in the PDHA1 gene are typically associated with a wide range of clinical presentations.S16 S17 Most patients have severe, often lethal early encephalopathy with lactic acidosis. Some cases have more chronic or subacute neurodegenerative disorders ranging from Leigh syndrome, episodes of ataxia or recurrent acute flaccid paralysis.S16 S17 Interestingly, a subset of patients manifest paroxysmal dystonia, which can be either isolated or combined with the aforementioned phenotypes or other clinical signs/symptoms including hypotonia, epilepsy and neurodevelopmental delay.S16 S17 The paroxysmal dystonia might be brought on by prolonged exercise, thus meeting the criteria for PED, or without any clear trigger, thus falling into the PNKD category.31 32 S17 S18 Attacks are sometimes reported to be hemidystonic. Similar attacks have been reported in patients with PDHX and DLAT mutations32 S16 S18 acknowledging that the latter two conditions are far less common than PDHA1 mutations.

Raised (serum or cerebrospinal fluid (CSF)) lactate and/or pyruvate levels along with pallidal hyperintensity suggesting striatal necrosis are important clues to suspect PDC deficiency mutations,S16-S18 but it is important to recognise that these might be lacking and, therefore, this condition should be considered in the differential diagnosis of isolated PED/PNKD even in the absence of any detectable biochemical or imaging abnormality, as it is a treatable condition that responds to thiamine supplementation.S19 In other cases, beneficial outcomes have been also reported with a ketogenic diet,S17 which supports energy failure as the main pathophysiological mechanism for the PxD occurring in the context of PDC deficiency.


GCH1 codes for the GTP cyclohydrolase I, a rate-limiting enzyme in the synthesis of tetrahydrobiopterin from GTP, mutations of which account for about 50% of dopa-responsive dystonias. A few patients with GCH1 mutations have been described with a phenotype consistent with PED. In 2010, Dale and colleagues described a family with two affected members with isolated PED.33 Attack duration was about 5 min and they never occurred at rest or during movement initiation.33 Moreover, Erro et al found two GCH1 carriers in a series of 16 patients with PED (12.5%).34 The phenomenology was that of dystonia and attacks were localised to the lower limbs.34 As expected, patients had a dramatic benefit on levodopa supplementation.34


Recently mutations in SCN8A, which encodes for sodium voltage-gated channel alpha subunit 8, have been suggested to be an alternative cause of the ICCA syndrome (ie, PKD and infantile convulsions) in one recent report.35 However, this proposal has been subsequently questionedS20 based on the evidence that, in one affected case, a ‘PKD‘ spell was recorded by video-electroencephalography and a cortical signal was documented, suggesting that such attacks might in fact be epileptic in nature. Therefore, it is unclear whether mutations in this gene truly cause PKD or not. However, it has to be acknowledged that in other reports SCN8A mutations have been associated with episodic dystonia, although the term paroxysmal dyskinesia was not explicitly used.36 As such, it is worth considering this condition in the differential diagnosis of PxD, especially when associated with epileptic seizures, particularly those refractory to antiepileptic therapy, and/or with neurodevelopmental delay.36


Mutations in ADCY5, which encodes for the adenylate cyclase 5, have been reported to cause a spectrum of (non-paroxysmal) movement disorders ranging from dystonia to chorea, sometimes associated with axial hypotonia and PxD.37–39 PxD do not always fit clearly within previously identified PxD categories and might be painful, a point of difference from PxD due to other genetic causes of PxD.37 39 Moreover, ADCY5-PxD may manifest, even within the same patient, as multiple subtypes, including PKD and PNKD.37 39 Of note, also at variance with other genetic disorders that can produce PxD, ADCY5 carriers characteristically develop PxD during sleep.39 Night-time dyskinesias (along with non-paroxysmal movement disorders) are therefore a clue to suspect ADCY5 mutations. Interestingly, Westenberger and colleagues have recently reported two unrelated ADCY5 patients with attacks reminiscent of alternating hemiplegia of childhood (AHC; see below) in the context of a more complex neurological picture including dysarthria, hypotonia and non-paroxysmal choreodystonia.40 Partial benefit has been reported with both tetrabenazineS21 and deep brain stimulation.S22


ATP1A3 mutations cause different clinical syndromes including AHC, rapid-onset dystonia parkinsonism and cerebellar ataxia with pes cavus and optic neuropathy, although there is increasing evidence of overlapping phenotypes.41 42 S23 S24 In the context of this review, we will only cover AHC. It is a largely sporadic disorder with onset within the first 18 months, by definition.41 42 Despite its name, the highly distinguished feature of AHC is occurrence of frequent episodes of either hemidystonia or hemiplegia, which can manifest together with other paroxysmal neurological signs including nystagmus, anarthria, dysphagia and seizures.41 42 S24 Duration of attacks ranges from a few minutes to several days, and episodes occur from repeatedly within a day to several times a month.41 42 S24 Attacks are almost invariably induced by emotional stressors, such as excitement or less frequently by physical stressors, including hypothermia or hyperthermia, respiratory tract infections and bright light.41 42 S24 Characteristically, there is a rostrocaudal gradient in the hemiplegic/hemidystonic episodes (face/neck>arm>leg)41 42 S24 which can aid the differentiation from other types of hemidystonic attacks. Hemiplegic and hemidystonic episodes typically shift from one side of the body to the other and typically disappear falling asleep.41 42 S24 Almost invariably the hemidystonic attacks are combined with other (interictal) features such as developmental impairment, walking difficulties/ataxia, muscular hypotonia, dysarthria and choreoathetosis. The mainstay of treatment is flunarizine (10–20 mg/day) as a prophylactic drug along with avoiding trigger situations.41 42 S24 Patients should be encouraged to sleep when attacks begin, using fast-acting benzodiazepines if necessary.


Mutations in the CACNA1A gene, which encodes for the calcium voltage-gated channel subunit alpha1 A, are associated with a number of phenotypes including SCA6, episodic ataxia type 2 as well as familial hemiplegic migraine.S25 In a minority of cases, CACNA1A mutations have been suggested to account for some cases of BPTI.43 44 It is characterised by episodes of head tilt with onset within the first 18 months of life that usually resolve by age 5.43 44 The attack duration ranges from 10 min to several days and is frequently accompanied with vomiting, pallor and ataxia.43 44 BPTI usually resolves after infancy, but can be replaced by paroxysmal vertigo and/or migraine.43 44 The co-occurrence of episodic ataxia, hemiplegic migraine and paroxysmal tonic upgaze in a single subject or in the family makes mutations in this gene more likely.45 This condition is self-limiting and usually no treatment is required.


SLC16A2 encodes for the monocarboxylate transporter type 8 (MCT8), which is required for transmembrane uptake of free triiodothyronine (fT3) from blood into neurons. MCT8 deficiency results in a complex, X linked disorder (also known as Allan-Herndon-Dudley syndrome) characterised by proximal hypotonia with poor head control, generalised muscular hypotrophy, microcephaly and marked developmental delay.46

The disorder is progressive and spasticity, ataxia and severe dysarthria complicate the clinical phenotype. In a subset of patients, a specific sort of PKD is observed.46 47 Attacks are in fact classically triggered by passive movements such as changing of their clothes or nappies or by lifting the children from one place to another.46 47 However, attacks can also be triggered by excitement, happiness or crying.46 47 Attacks are brief, lasting seconds to few minutes, and are dystonic in nature.46 47 The hallmark of MCT8 deficiency is raised serum concentration of fT3.46

Other causes of PxD

Table 2 lists the conditions that have been associated with PxD or similar episodes of choreodystonia.13 48 S26-S55 These include a variety of acquired, immunological and neurodegenerative causes that were formerly ascribed to secondary PxD.13 48 For this reason, we have also included here brain calcification, by virtue of the fact that a lesional mechanism is assumed in such cases.S35 S36 However, two PxD families have been recently reported on, in whom genetic analysis revealed novel mutations in SLC20A2 and PDGFB genes, respectively.S48 S49 The fact that all affected members shared the phenotype of isolated PxD with normal interictal examinationS48 S49might support the idea that PxD are intrinsically associated with these mutations rather than being merely secondary to basal ganglia calcification, as assumed in earlier reports of the pregenetic era.S49 This, however, remains speculative, but it further exemplifies the ambiguity regarding the concept of primary and secondary PxD.

In general, PxD due to acquired, immunological or neurodegenerative causes present usually at a later age compared with the main genetic forms reviewed above and manifest with additional signs or symptoms that will easily drive the diagnostic workup and lead to the correct diagnosis and appropriate management in most cases.

Recently, DEPDC5 S10 and CHRNA4 49 mutations have been associated with the syndrome of PKD plus epilepsy in single families. However, these two genes are also a cause of ADNFLE50 and it remains to be seen whether these episodes of paroxysmal dystonia are epileptic in nature or not. Moreover, these results require replication before screening of these genes might be recommended in clinical practice.

Table 2

Different aetiologies associated with episodic movement disorders resembling paroxysmal dyskinesias

Proposed criteria for case definition and diagnostic strategy

PxD represent clinical syndromes where the disorder of movement is intermittent in nature (and thus does not encompass exacerbation of existing abnormalities). The intermittent character of the disorder means it is not continuous or steady, but should not be used to refer to disorders that wax-and-wane over a period of time such as tics.

As to the phenomenology, the clinical spectrum ranges from dystonia to chorea, with ballism being possible but does not encompass tremor or myoclonus. Such a clarification automatically excludes stimulus-sensitive myoclonus or startle syndromes from this category. Moreover, excluded from this definition are those phenomena that are clearly drug induced (ie, acute dystonic reaction or levodopa-induced dyskinesias in the context of Parkinson's disease, for instance).

Using this definition, the first step for the differential diagnosis is to decide whether the clinical abnormality is in fact a PxD or not. Epilepsy, tonic spasms, tetany, neuromyotonia, periodic paralyses and episodic ataxias, all of which can produce intermittent disordered movements, need to be excluded clinically and/or by ancillary investigations whenever appropriate. Moreover, psychogenic/functional causes have to be ruled out. While some authors have suggested that the diagnosis of psychogenic/functional paroxysmal movement disorders is fundamentally one of exclusion, we would rather support alternative claims that the diagnosis should be based on the presence of positive signs: these include profound within-subject phenomenological variability with marked increases in attack frequency and severity during examination, highly variable attack duration, presence of several and non-specific triggers, frequent alteration of responsiveness during attacks, medically unexplained somatic or neurological symptoms and, finally, atypical response to medications.51 52 These clues will make a positive diagnosis of psychogenic/functional movement disorders likely, without the need for additional investigations in most, if not all, cases.

Once the clinical syndrome of PxD is established, the second step is to fully characterise the attacks in terms of trigger and duration, further exploring the presence of family history and additional clinical features (by history or on examination) and to set the identified clinical syndrome in the context of age at onset. A clinical syndrome with onset in childhood, which is characterised by attacks with specific triggers and duration, is most likely to be genetic in nature. The definition of the trigger(s), duration and body distribution of the attacks, as well as the presence of suggestive associated clinical features and the pattern of inheritance should help the clinician to drive the genetic analysis (figure 1). Most of these disorders account for those forms that were formerly considered primary PxD and include PKD, PNKD and PED. While supporting the trigger-based approach as very useful, we further suggest some modifications and clarifications.

Figure 1

Suggested genetic mutations that should be tested for according to clinical presentations. ^Mimics include epilepsy, tonic spasms, tetany, neuromyotonia, periodic paralyses and episodic ataxias. #These might include stress, anxiety, fatigue, fasting, startle and sleep deprivation. §By history or on examination. −, none/negative; +, present; ≅, might have a kinesigenic trigger; *Including 16p11.2 deletion; **Should include glucose, lactate, pterins, and dopamine metabolites; A, axtaxia; AD, autosomal dominant; Au, autonomic dysfunction; E, epilepsy; EA, episodic ataxia; M, migraine; HP, hemiplegia; MR, mental retardation; MD, non-paroxysmal movement disorders; S, spasticity.

For instance, the term non-kinesigenic does not carry any useful information rather than specifying that the trigger is not kinesigenic. It reflects the absence, rather than the presence, of a clinical feature with the obvious implication that, with one notable exception, the majority of non-kinesigenic triggers are non-specific and are shared across different PxD subtypes, being therefore not predictive of the underlying genetic defect. The exception is represented by alcohol/caffeine in the case of MR-1 mutations. The presence of this trigger is highly specific of MR-1 mutations, being present in about 95% of carriers. We therefore advocate considering alcohol/caffeine sensitivity as a distinctive trigger for PxD rather than relegating it within the (unspecific) subgroup of non-kinesigenic triggers (figure 1). The vagueness of the non-kinesigenic category further justifies the fact that, in some instances, other clinical features such as the body distribution of the attacks might prevail in the definition of the paroxysmal movement disorders. This is the case of dystonic and hemiplegic attacks alternating from one body side to the other (AHC) or episodic attacks of neck dystonia (BPTI).

Moreover, we support the concept of PxD occurring during sleep as a further trigger-based subtype. Although sleep is not strictly a trigger, it can be considered equivalent to other triggers since they all give an answer to the question of when PxD occur (ie, after sudden movements, after alcohol/caffeine ingestion, during sleep, and so on).

At variance with former classifications of PxD, we discard the criterion of normal interictal examination. As such, all the aforementioned trigger-based subtypes can be isolated or combined with additional features. This reiterates a previous suggestion proposing the stratification of PxD into ‘pure’ and ‘complicated’ forms,53 based on the absence or presence of additional interictal neurological signs, respectively. We also advocate this approach for two main reasons. First, there is evidence of clinical heterogeneity for single gene mutations. For instance, PRRT2 and SLC2A1 mutations can produce either isolated or combined PxD. Second, we believe the syndromic approach will facilitate the differential diagnosis (figure 1). For instance, the presence of epilepsy in a patient with ‘classic’ PNKD will make the presence of MR-1 mutations very unlikely. Such a syndromic approach would also prioritise some investigations over others. This might be the case for PED in which CSF examination for glucose, pterin pathway components, pyruvate and lactate would provide more information than imaging.

Whenever the attacks lack specific and consistent triggers, have variable duration and when the onset is in adulthood, ‘symptomatic’ causes (table 2) need to be excluded, especially in the absence of family history. Given the huge variability of symptomatic PxD forms in their clinical presentation, even within the single subject, it is hard to identify any phenotypic patterns suggestive of the underlying aetiology. In general, certain findings including painful attacks, fluctuating levels of consciousness and dysautonomic crises that are not classically seen in the genetic conditions reviewed here should prompt the clinician to rule out acquired causes. In such cases, an initial diagnostic workup including metabolic and electrolyte panels, investigation for autoimmune disorders and brain imaging will allow a definitive diagnosis to be determined in most cases.


In recent years, great advances have led to a better understanding of the broad spectrum of genetic conditions underlying the PxD. This has increasingly challenged the former phenomenological classification as well as the idea that any specific phenotypes were associated to single gene mutations. Such an argument reiterates previous proposals14 that classification of PxD should follow a two-pronged method, according to which both the clinical phenotype and the specific genetic mutation should be stated (i.e., PRRT2-PKD).

It is worth specifying that our proposal does not represent a classification scheme, but reflects an algorithmic approach to help clinicians in the differential diagnosis. First, it gives clarity to the definition of PxD. Second, it has the merit of encompassing an increasing number of recently identified conditions with PxD as a feature that would escape the current classification. On the other hand, the phenotypic and genetic heterogeneity of PxD highlighted here might render the test of candidate genes, based on a specific clinical syndrome, unsuccessful. In this context, it might be argued that next generation sequencing approaches would better apply to the need of a rapid comprehensive genetic screening.54 This would further reduce costs in comparison to single gene testing.55

Of course there will be cases where no definitive cause for PxD can be found and treatment is to be pursued empirically. These can be labelled as idiopathic forms while awaiting for further elucidation of genetic or other causes. In turn, our proposal will require updating as soon as novel evidence is available. Additional references can be found in the online supplementary file 1.

Supplementary data


We thank Dr K Bertram for having edited the text.


View Abstract


  • Contributors RE: conception, writing the first draft. KPB: conception, reviewing the draft.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests RE received honoraria for speaking at meetings from TEVA, ZAMBON and the International Parkinson’s Disease and Movement Disorders Society. He receives royalties from publication of Case Studies in Movement Disorders–Common and uncommon presentations (Cambridge University Press, 2017). KPB has received grant support from Welcome/MRC, NIHR, Parkinsons’s UK and EU Horizon 2020. He receives royalties from publication of the Oxford Specialist Handbook Parkinson's Disease and Other Movement Disorders (Oxford University Press, 2008), of Marsden's Book of Movement Disorders (Oxford University Press, 2012), and of Case Studies in Movement Disorders–Common and uncommon presentations (Cambridge University Press, 2017). He has received honoraria/personal compensation for participating as consultant/scientific board member from Ipsen, Allergan, Merz and honoraria for speaking at meetings and from Allergan, Ipsen, Merz, Sun Pharma, Teva, UCB Pharmaceuticals and from the American Academy of Neurology and the International Parkinson’s Disease and Movement Disorders Society.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.