Dopamine

The first and most persistent hypothesis of tardive dyskinesia aetiology relates to postsynaptic dopamine (D₂) receptor supersensitivity. This was proposed independently in 1970 by neurologist Harold Klawans and Nobel Laureate pharmacologist Arvid Carlsson, by analogy with the outwardly similar abnormalities evident in people taking long-term L-dopa for parkinson's disease. The idea is that chronic antagonist blockade results in increased postsynaptic dopamine receptor numbers (and/or sensitivity of unblocked receptors), resulting, in line with Cannon's law of denervation, in a supersensitised motor system. In functional terms, this would place tardive dyskinesia as the pathophysiological ‘opposite’ of parkinsonism – domination of striatal dopaminergic activity rather than its increasing subtraction (Fig. 1). The promoting factor may be dopamine excess but the symptom mediator is cholinergic insufficiency.

FIG 1 Tardive dyskinesia: schematic representation of functional implications of the postsynaptic receptor supersensitivity hypothesis. From Owens (Reference Owens2014).

The value of the theory is that it explains the phenomenon of ‘suppression’, where even quite marked degrees of tardive dyskinesia can be, at least temporarily, ameliorated or abolished by antipsychotic dose escalation. However, cross-sectional comparisons of tardive dyskinesia and L-dopa-induced dyskinesias reveal similarity, not identity (Karson Reference Karson, Jeste and LeWitt1983), and this theory raises more questions than it answers (Owens Reference Owens2014). At best, it should be seen as providing limited insight.

Beyond dopamine

Which means coming up with an alternative! Although several have been proposed, thus far none provides compelling understanding.

1. 1 Presynaptic dopaminergic/noradrenergic hyperactivity hypothesis (Jeste Reference Jeste and Wyatt1981): an early and neglected challenge to conventional wisdom came when emphasis was first shifted to presynaptic mechanisms, with the proposal that presynaptic dopaminergic overactivity and noradrenergic hyperactivity were together necessary for at least some types of tardive dyskinesia. Beta-blockers provide a ready test of this hypothesis but have been surprisingly neglected as potential therapeutic options, although interest is reviving (Hatcher-Martin Reference Hatcher-Martin, Armstrong and Scorr2016). This is a theory whose implications are now topical but remain to be exploited.

2. 2 Cholinergic interneuron burnout hypothesis (Miller Reference Miller and Chouinard1993): acetylcholine plays a crucial role in striatal function, the striatum being the source of the highest levels of acetylcholine markers (transmitter plus synthesising/metabolising enzymes) in the brain. Most of this acetylcholine is associated with cholinergic interneurons, although these account for only 1–2% of striatal cell numbers, a feat achieved through extreme arborisation. Under physiological conditions, cholinergic interneurons tonically discharge, a process modulated by dopaminergic inhibition. The proposal is that with long-term dopamine blockade, cholinergic interneurons, in effect, burn out from loss of inhibitory control. This hypothesis was an interesting departure from the sterility of supersensitivity, not least because it opens up for consideration a wide range of other transmitter systems that themselves interact with cholinergic interneurons, such as GABAergic (Fibiger Reference Fibiger and Lloyd1984) and glutamatergic, both ionotropic and metabotropic (Lim Reference Lim, Kang and McGehee2014). However, as with supersensitivity, confirmation is lacking.

3. 3 Excitatory/oxidative stress hypothesis (Lohr Reference Lohr1991): the notion that disease can be mediated via ultra-brief, highly reactive chemical moieties is well-rooted in pathology. In relation to tardive dyskinesia, the proposal is that increased presynaptic dopamine activity (turnover), initially identified with postsynaptic antagonism, creates damaging dopamine quinones, and possibly hydrogen peroxide, through the action of monoamine oxidase. A further suggestion is that damage emanates from deficient production of free-radical defences, such as superoxidase dismutase (Cadet Reference Cadet, Lohr and Jeste1986), and increased lipid peroxidation, which result in membrane damage/dysfunction and possible cell death (Elkashef Reference Elkashef and Wyatt1999). Whatever the mechanism, the consequence is degeneration of dopamine-receptor-rich, GABAergic striatal interneurons, themselves normally inhibitory. A degree of experimental evidence supports this hypothesis in, for example, depositions of heavy metals and elevated cerebrospinal fluid (CSF) lipid peroxidase parameters but, again, it has not matured beyond a theoretical proposal. A major impediment has been the generally disappointing results of antioxidants as treatments.

4. 4 Synaptic plasticity hypothesis (Teo Reference Teo, Edwards and Bhatia2012): this taps into a topical field in neuroscience and, according to its proposers, could provide the missing link between supersensitivity and interneuron dysfunction. It sets out a number of possibilities, one of which is that D₂ hypersensitisation (of striatal medium spiny neurons and feedforward fast-spiking interneurons) affects the plasticity of glutamatergic interneuron synapses, resulting in disordered striato–thalamo–cortical output to the sensorimotor cortex, one consequence being reduced activity in cortical inhibitory circuits, with increasing excitability in executive motor areas. Such miscoding of motor programmes could explain persistence of disorder after removal of the causative antipsychotic. This elegant theory is in line with modern proposals on the pathophysiology of other hyperkinetic states, especially chronic dystonia (Calabresi Reference Calabresi, Pisani and Rothwell2016), and may have its greatest relevance to disorder of this type but remains another hypothesis without validation.

It is possible each of these broad brushes paints part of the picture but none, as yet, allows us to make that crucial leap to innovative therapy.

Primary prevention

In the absence of treatments with authentic antidyskinetic actions, the management of tardive dyskinesia remains firmly rooted in primary prevention – limiting indications, limiting doses and avoiding scenarios of heightened risk (Box 2).

Antipsychotic prescribing has increased fourfold in the past two decades, mainly from a rise in the use of newer compounds for an expanding catalogue of licensed indications (Verdoux Reference Verdoux, Tournier and Begaud2010). It is essential to appreciate that ‘licensing’ is a commercial imperative, where companies (‘marketing authorisation holders’) are required to present efficacy and safety data to support product promotion. It does not mean that licensed products are the only drugs for a particular job – nor that a particular ‘job’ requires a product at all! Antipsychotics, as the name implies, should be restricted to ‘psychotic’ indications and not used for management of non-specific problems such as insomnia or disruptive behaviours, unless in extremis and only when all other pharmacological and psychological routes have been exhausted. It was only in the 1970s that the original inference of ‘psychosis’ (‘severity’) was supplanted by a concept based on the presence of particular symptoms (hallucinations, delusions, thought disorder and sometimes affective incongruity). In clinical practice, ‘antipsychotics’ would be appropriately indicated in mania/hypomania, traditionally considered ‘psychotic’ disorders.

Minimum effective dosing is the key feature of antipsychotic use but, because of pharmacokinetic diversity, is difficult to standardise across patients. The physician's skill is in tailoring therapy to individual need by choice of compound and dose, both initial and incremental, a key element of which is adopting realistic timescales to allow for adaptation of biological systems. Pragmatic trials have consistently failed to support the view that newer antipsychotics have an inherently reduced liability to promote extrapyramidal side-effects (EPS) (Lieberman Reference Lieberman, Stroup and McEvoy2005; Jones Reference Jones, Barnes and Davies2006; Fischer-Barnicol Reference Fischer-Barnicol, Lanquillon and Haen2008), leading to the inevitable conclusion that their advantage lies in licensed doses, pharmacokinetically derived, being lower than the empirically derived doses traditionally utilised with older drugs (Owens Reference Owens2014). One might postulate that the comparability of tardive dyskinesia rates that has emerged with longer-term use of newer compounds relates to the fact that, in practice, doses of these tended to rise post-launch (Citrome Reference Citrome, Jaffe and Levine2005, Reference Citrome and Kantrowitz2009).

Risky prescribing cannot always be avoided but the prescriber's responsibility is to be aware of variable risk. For example, the ageing brain does not tolerate antipsychotics well, and utilising this class in older patients, especially with first exposure, needs careful thought.

Pharmacological treatment

Even with diligent care, individual predisposition dictates that some patients will still develop disorder but, as a result of multifarious and sometimes competing confounds in inadequately powered studies, the research literature on their treatment is of ‘poor’ or ‘very poor’ quality (Bergman Reference Bergman, Walker and Nikolakopoulou2017) and it is not possible to specify a definitive algorithm.

Assessing the efficacy of treatments for tardive dyskinesia is uniquely complex (Box 3). The condition itself comprises different types of disorder (Owens Reference Owens2018), which may be differentially affected by therapeutic interventions, and a large part of its clinical presentation is exquisitely sensitive to modifiers, both external and internal (environment, mental state, etc.). As a result, tardive dyskinesia is highly variable over time, even within the day. Coincidental medication can have an apparently favourable effect but for entirely non-specific reasons. There are issues surrounding the recording instrument chosen, how comprehensive the assessment is and how data are analysed. Importantly, there issues about how one militates against unconscious biases, which extends beyond just treatment allocation (see below: critique). Not least is power. Perhaps the ideal study is un-doable but in appraising those that do come to publication, it is important to be aware of these complexities.

Antipsychotic dose increment

It has long been known that the most effective management of tardive dyskinesia is to increase antipsychotic dose (Jeste Reference Jeste, Lohr and Clark1988) (Table 1). This apparently contrary observation reflects the phenomenon of suppression, thought to result from further blockade of oversensitised dopamine receptors, or the related masking by emergent parkinsonism. It is a likely explanation for much ‘improvement’ in switching studies where the switch is to a more potent compound.

TABLE 1 Average rates of improvement in tardive dyskinesia in controlled treatment trials

Source: Owens (Reference Owens2014); data from Jeste et al (Reference Jeste, Lohr and Clark1988).

Never popular, such strategies remain largely of theoretical interest and should only be considered in tertiary specialist centres.

Presynaptic depleters

The surge of publications on tardive dyskinesia in recent years has been driven by the fact that two variants on this theme (valbenazine and deutetrabenazine: see below) have received US licenses. It is important to see these developments in context and to evaluate what is on offer hard-headedly.

Relevant neurophysiology

Recycled neurotransmitters are taken up across the presynaptic membrane by specific protein transporters, and stored, as is newly synthesised transmitter, in vesicles, which are mainly formed by budding from endoplasmic reticulum (Fig. 2). Vesicular monoamine transporter 2 (VMAT2) is a glycoprotein with 12 transmembrane domains that acts to facilitate the uptake of cytosolic transmitter into these storage vesicles to await use/reuse.Footnote ^a This is an active, energy-dependent process involving one of the widely expressed adenosine triphosphatase (ATPase) enzymes, vacuolar-type H-ATPase (V-ATPase) which, by pumping protons into newly formed vesicles, produces a conformational change in the transporter, allowing importation of transmitter. Excess transmitter left in the cytosol is broken down by monoamine oxidase. VMAT2 is non-selective and deals with re-storage of serotonin and histamine as well as the catecholamines, dopamine and noradrenaline.

FIG 2 Storage vesicles: life-cycle and pharmacological modification. After Arbuthnott & and Garcia-Munoz (Reference Arbuthnott, Garcia-Munoz, Johnstone, Owens and Lawrie2010).

Vesicles are docked to the internal presynaptic membrane, most commonly by synchronous fusion, an immensely complex process involving recruitment of a series of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins and binding partners, where docking is precisely time-locked to terminal depolarisation (Li Reference Li and Kavalali2017). Transmitter release, by a process called exocytosis, follows influx of calcium, with docked vesicles releasing their content into the synaptic cleft either through generation of a fusion pore or by complete membrane fusion and rupture (Fig. 2). In most instances, the ‘empty’ vesicle is incorporated into the presynaptic membrane by collapse fusion. The presynaptic membrane also contains pits coated with the protein clathrin, which provide a source of new vesicle formation. Although incredibly complex, the vesicular cycle, from budding to fusion/collapse, can be completed within a minute (Li Reference Li and Kavalali2017).

The precision of synaptic messaging is such that variations in presynaptic release differentially affect downstream receptor activity almost instantaneously. In addition to theoretically addressing disease processes at origin, presynaptic depleters, mediating reduced transmitter release, represent a gentler, more subtle way of modulating synaptic events than ‘heavy-handed’ postsynaptic receptor antagonism.

Clinical pharmacology

The traditional approach: tetrabenazine

Extracts of Rauwolfia serpentina (Indian snakeroot) have been used in Ayurvedic medicine for centuries, and one of the first compounds to be refined from the plant, reserpine, was also one of the first drugs of the psychopharmacology age. Reserpine interacts with VMAT1 but also binds irreversibly to its own VMAT2 binding site. The drug fell from widespread psychiatric use because of poor tolerability. However, a synthetic analogue, tetrabenazine, which binds selectively to VMAT2 and reversibly to a separate, or possibly overlapping, site from reserpine (Grigoriadis Reference Grigoriadis, Smith and Hoare2017), was promoted as an antipsychotic in the UK from 1958 until its withdrawal in 1966 on a reputation of poor efficacy. It did, however, remain on the market in a number of European countries, including the UK, as a treatment for hyperkinetic movement disorders, including tardive dyskinesia. In the USA, approval was only obtained in 2008 and then for Huntington's disease alone – but with an unusual restriction. In addition to a black box warning in relation to depression and suicidality, the Food and Drug Administration (FDA) imposed the recommendation for CYP2D6 genotyping with higher doses. This might be seen as running ahead of the science and as contributing to an unnecessarily dark reputation. This difference in availability and perception between Europe and the USA is important in assessing how great a leap forward new compounds might be perceived as representing.

Although better tolerated than reserpine, tetrabenazine still has a prominent side-effect profile (Guay Reference Guay2010), some of which relates to the fact that several metabolites are active D₁ and D₂ antagonists (Grigoriadis Reference Grigoriadis, Smith and Hoare2017). Indeed, its tolerability profile is very suggestive of a mild, sedative antipsychotic (Table 2) – although presentation of adverse effects by patient report can be misleading: for example, ‘insomnia’ in a prominently sedative D₂-modifying drug such as tetrabenazine is likely to reflect akathisia. In addition to patient-reported effects, tetrabenazine has a slight ability to prolong the QTc interval (according to the Summary of Product Characteristics by, on average, 8 ms) and caution is advised when co-prescribing potent CYP2D6 inhibitors. With a short half-life (approximately 2–3 h) and rapid and comprehensive brain penetrance but equally rapid egress, it requires dosing twice or, preferably, three times a day. Tetrabenazine has never been widely used in UK psychiatry, remaining a niche neurological product, and perhaps as a result is costly for a compound that has been around for so long.

TABLE 2 Adverse effects^a of tetrabenazine (>10%)

a Only those reported by >10% of patients are shown here.

The new approach: valbenazine and deutetrabenazine

Two approaches attempt to address the tolerability problems of tetrabenazine by modifying kinetics.

Tetrabenazine undergoes heavy first-pass metabolism, via mainly CYP2D6 (with 1A2 a minor pathway), resulting in bioavailability of less than 5%. From a complex range of outcomes, four discrete isomeric secondary alcohols are created, collectively referred to as dihydrotetrabenazine (Grigoriadis Reference Grigoriadis, Smith and Hoare2017). All four inhibit VMAT2 but one (R,R,R-HTBZ), the most potent inhibitor, has been esterified with the amino acid L-valine to produce valbenazine. The (R,R,R)-isomer demonstrates high specificity for the VMAT2 transporter while being virtually devoid of receptor interactions (Grigoriadis Reference Grigoriadis, Smith and Hoare2017).

Valbenazine acts as a prodrug which, following absorption, releases active HTBZ by hydrolysis to inhibit VMAT2. Amino acid esterification is increasingly implemented to improve the kinetic qualities of drugs of short half-life or with extensive first-pass, which require repeat dosing with resulting increases in peak-level side-effects (c.f. lisdexamfetamine).

The second method for modifying kinetics adopts a different, but again increasingly utilised, approach – deuteration. This involves replacing hydrogen atoms (in this case six), with deuterium, also known as heavy hydrogen (D or ²H), the resulting compound being referred to as deutetrabenazine. Deuteration slows a drug's metabolism without conveying other significant chemical changes.

Efficacy/tolerability of valbenazine and deutetrabenazine

Both valbenazine and deutetrabenazine have been shown to be effective in significantly reducing global tardive dyskinesia ratings using, as the primary outcome, total scores on the AIMS (O’Brien Reference O'Brien, Jiminez and Hauser2015; Anderson Reference Anderson, Stamler and David2017; Fernandez Reference Fernandez, Factor and Hauser2017; Hauser Reference Hauser, Factor and Marder2017), with 40% and 33% on the higher valbenazine (80 mg/day) and deutetrabenazine (36 mg/day) doses respectively achieving 50% or greater reduction in AIMS totals. Furthermore, both compounds are well tolerated (Table 3), although it is important to appreciate that adverse effects profiles, as presented in Tables 2 and 3, permit only relative, and not absolute, tolerability comparisons. Nonetheless, the likelihood of being helped or harmed (LHH), a ratio of the number needed to treat (NNT) to achieve a 50% reduction in AIMS totals over the number needed to harm (NNH) from discontinuation because of an adverse event, has been calculated at 15.2 for valbenazine and 27 for deutetrabenazine (Citrome Reference Citrome2017a,Reference Citromeb), projecting both drugs as highly favourable therapeutic agents.

TABLE 3 Adverse effects of valbenazine and deutetrabenazine

a. Data from Hauser et al (Reference Hauser, Factor and Marder2017).

b. Data from Anderson et al, Reference Anderson, Stamler and David2017).

c. Data from Fernandez et al (Reference Fernandez, Factor and Hauser2017).

Critique

Studies evaluating these compounds, which relate to short-term efficacy (valbenazine 6 weeks; deutetrabenazine 12 weeks), are of good quality (Davis Reference Davis, Miller and Kalsi2017), addressing a number of problems that so confounded past work, in particular the complex issue of masking (Box 3). This involves not just unawareness of treatment allocation but must take into account sequencing or expectancy bias (expectation of improvement as a study progresses) and the phenomenon of prior context (overflow of previous to current assessment), among other biases. Even though these issues were highlighted over two decades ago (Owens Reference Owens, Finlayson and Mercer1997) these are the first neuropharmacology arenas, to the author's knowledge, in which trial design included randomised masked video ratings. Both these programmes therefore give a major nod to point 11 in Box 3. By adopting a global impression recording instrument and presenting data as total scores, neither programme allows for judgements on what type of movement disorder – if any – is responding preferentially or in which body part it occurs, although further detail may emerge. Participants did partially fulfil the stability criterion in that they were required to have been on antidopaminergic medication for 3 months or more (>1 month in over-60s for deutetrabenazine) but regimes were only required to be stable for 30 days and additional medications, although discouraged, were permitted.

The profile of presynaptic depleters is likely to rise as they represent neglected drug targets for CNS disorders, especially in relation to schizophrenia, where their pharmacology fits well with the theoretical shift in the dopamine hypothesis from its original postsynaptic receptor focus to more recent emphasis on presynaptic mechanisms (Howes Reference Howes and Kapur2009). However, in fully evaluating these drugs – and justifying costs – we must consider research findings critically. In relation to tardive dyskinesia, a condition so readily modified by mental state and environment, an important confound relates to point 6 in Box 3 – the nature of the placebo. The question is the extent to which any benefits can be attributed to specific antidyskinetic pharmacological action(s), or result from non-specific factors such as sedation or suppression.

Tetrabenazine is quite markedly sedative (Table 2) but creating a pro-drug analogue or slowing kinetics appears to reduce this problem (Table 3). It is, however, difficult to assess the extent of this benefit, as direct head-to-head comparisons have not been undertaken. Data, as noted, only allow for relative not absolute conclusions, an important distinction (Rodrigues Reference Rodrigues, Duarte and Costa2017). Although well tolerated, both valbenazine and deutetrabenazine do have sufficient relevant actions to contribute to, if not account for, the relatively modest reductions in AIMS ratings reported.

Since deuteration favourably extends kinetics but does not fundamentally alter metabolism, one must, furthermore, assume that deutetrabenazine is associated with production of D₂-antagonistic metabolites, opening another possible window on reduced AIMS ratings. Although parkinsonism was assessed in the valbenazine study using the much criticised Simpson–Angus Scale (SAS) (Owens Reference Owens2014), both deutetrabenazine studies used the motor subscales of the Unified Parkinson's Disease Rating Scale (UPDRS), a comprehensive and reliable instrument (Owens Reference Owens2014). Neither compounds were associated with significant increases in Parkinsonian ratings. This, however, does not mean that subthreshold increases in D₂ antagonism are not relevant to the drug's action. Caution is supported by the fact that Anderson et al (Reference Anderson, Stamler and David2017) found that improvements on deutetrabenazine were greater in those not receiving a concomitant antipsychotic, compatible with the view that some benefit emanates from suppression.

In general, such comments might seem trifling if efficacy is striking. Presentation of trial results in terms of percentages achieving 50% reduction in signs certainly creates an impact but the actual drug-attributable reductions in both the valbenazine and deutetrabenazine studies were modest. On a scale with a possible total score of 21, reductions relative to placebo were, on average, 3.1 AIMS scale points for valbenazine 80 mg/day (1.8 for the 40 mg/day dose) (Hauser Reference Hauser, Factor and Marder2017) and less than 2 AIMS points for the two higher deutetrabenazine doses (Anderson Reference Anderson, Stamler and David2017; Fernandez Reference Fernandez, Factor and Hauser2017). Variance was, as would be expected, considerable. Although the video ratings showed statistically significant effects, the Clinical Global Impression ratings were not consistently significantly different (Fernandez Reference Fernandez, Factor and Hauser2017; Hauser Reference Hauser, Factor and Marder2017). This means that on-site clinicians dealing with the patients regularly could not necessarily detect the improvement trend. This is important, bearing in mind the biases, noted above, to which such assessors are prone.

Thus, there is reason to believe that the evidence for valbenazine and deutetrabenazine suggests small therapeutic steps rather than large leaps forward and that, despite a known pharmacology, mode of benefit remains open to speculation. While any agent producing improvements in tardive dyskinesia is to be welcomed, the qualifications that must be imposed on bottom-line data are important, especially when one turns to perhaps the major issue in trial design and interpretation – power (Box 3). Although both trial programmes comprised impressively large samples in comparison to previous literature, it is questionable whether even they were adequately powered to address the primary question of efficacy, especially in expectation of slight change. One comprehensive review suggests that very large samples – in excess of 800 participants – would ideally be needed to evaluate properly drug interventions for tardive dyskinesia (Bergman Reference Bergman, Walker and Nikolakopoulou2017), implying sample sizes two to three times those for the pivotal valbenazine and deutetrabenazine studies. This may be impracticable and perhaps, with the key issue of masking systematically acknowledged, unnecessary but the point is crucial to interpretation and continues to support caution.