‘Fake evidence’ about tardive dyskinesia?

   Intended as a statement of consensus to inform a standard of care in persons with tardive dyskinesia, this report, by authors of record Stanley Caroff, Les Citrome, and colleagues, published last spring in the Journal of Clinical Psychiatry (2020, March-April (81(2): 19cs12983) is presented in the form of an investigational study, but it is not designed to disprove a null hypothesis, nor does it purport to gather systematic observations in support of a falsifiable hypothesis that answers a scientific question. In fact, it is not so much study as opportunity to endorse the recently launched product of its sponsor (valbenazine (Ingrezza) and Neurocrine, Inc., respectively) by including it as a first-line pharmacotherapy for tardive dyskinesia. This endorsement appears among several recommendations for which the sponsor and authors of record (so designated because their contributions are not attributed) seek the imprimatur conferred by inclusion in a consensus statement. In fact, there is no study to be found when no one who is not an employee of , contractor paid by, consultant to, or speaker for the manufacturer is credited with involvement in the study. By its own disclosure, the manufacturer conceived of, designed, hired a contractor to execute an uncontrolled study using a methodology better suited to identifying agreement rather than quality of evidence-based recommendations–a much- and long-criticized “Delphi method,”lacking random selection of sample, controls, blindedness, source of information for evaluating recommendations, and even lacking statistical analysis of data– which the manufacturer collected, collated, and analyzed “in-house” before hiring another contractor to write up the proceedings as a study and submit it for publication.

The participants were a non-randomly preselected group of thirty persons that a non-randomly selected 11-member “steering committee” recruited by virtue of their having met the steering committee’s impromptu criteria for expertise. The steering committe supplied participants with a “core bibliography” of tardive dyskinesia literature consisting of studies, reviews, and other, unspecified content published no earlier than 2007. Both the first round and the second (feedback) round were conducted by a contractor hired by the sponsor, and are not further described beyond disclosure that they were administered pursuant to a modified “Delphi method,” about which the author offers no details other than that the participants filled out a multiple-choice-like answer sheet following a statement about tardive dyskinesia; the answer sheets were then processed in an unspecified manner and used during a second round of non-verbal responses (to whom and from whom not disclosed) whose objective was to reconcile first-round information and yield greater agreement. How this was done is not described. The results, submitted in undisclosed format after the first round, were converted to scores on a “Likert scale” of numerical ratings accompanied by written descriptions from “strongly disagree” and “very unimportant” (the relationship between agreement and importance not clarified) Likert ratings from the second round were subtracted from those of the first round and analyzed by an unidentified statistical method. The results of the latter are not included in the exposition of the study. Of those results, the authors state that they were used for “descriptive purposes only” and go on to locate them in “Supplementary Tables 1-3, which are not included in the text (which includes only a “Figure 1” and three “boxes,” wherein no statistical results are given). Items on which fewer than 25 percent of participants agreed during the first round were excluded from the second round and from the data analysis. (The authors state that items about which 75 percent or more of the participants agree were similarly excluded, but they make reference later in their paper to several items with percentages of agreement exceeding 75 percent after the first round, and do not explain the apparent discrepancy.

   Little is presented with respect to results that pertain to consensus development, and what little there is does not include the quality of evidence supporting the recommendations, which were the de facto “findings” and included among two classes of pharmacotherapy recommended as “first-line” was the authors’ product, designated a VMAT2 inhibitor (for vesicular monoamine transporter protein, type 2). Tetrabenazine, of which valbenazine is a closely related derivative structurally and pharmacologically similar (with similar milligram-potency), whose relevance to the context of the “study” derives from the fact that is has been used clinically for hyperkinetic movement disorders since the 1960’s; in fact, a preponderance of literature pertaining to tetrabenazine was published before 2007–a bemusing coincidence, since the “bibliographic core” to which participants were restricted omitted all literature published before 2007 (without giving any reason for omission).

The results of the study are given in the form of the authors’ recommendations for screening, treatment, and monitoring persons with tardive dyskinesia.

The authors’ recommendations for screening differ from those previously published in two respects: One, wherein they assert that a diagnosis of tardive dyskinesia is “suggested” by a finding of “abnormal movement in any one body part,” but do not specify whether frowning or grimacing (both rateable on the Abnormal Involuntary Movement Scale and both occurring more often in contexts other than tardive dyskinesia) should be excluded from the AIMS scale out of concern that the latter might militate toward falsely positive diagnoses or falsely positive impressions of worsening in persons previously diagnosed with tardive dyskinesia. Moreover, the authors recommend that no longer than one month be allowed to elapse before persons started on a neuroleptic return for screening of tardive dyskinesia. They cite no reports of latency as brief as one month to onset of tardive dyskinesia, and do not cite prior studies indicating a shortest latency of three months. Previously identified risk factors that they leave out are diagnosis of mood disorder, chromosomal sex, and diabetes mellitus.

    The authors do not clarify the meaning of “risk factor” as an association of a diagnosis of tardive dyskinesia that is not necessarily causal and should not be construed to mean more than that the risk factor and the diagnosis of tardive dyskinesia occurred in the same person more often than would be expected by chance. Marijuana smoking, for example, may be identified as a risk factor of tardive dyskinesia because marijuana smoking is over-represented among persons with schizophrenia, as is neuroleptic exposure, which is causally correlated with development of tardive dyskinesia. That may be why marijuana smoking and use of other substances may be identified as risk factors for but not causes of tardive dyskinesia. More closely related causally is gynecomastia, not mentioned in the study but recently reported to be a risk factor for tardive dyskinesia. Gynecomastia is strongly correlated with hyperprolactinemia, which is associated causally with neuroleptic exposure (and, thereby, with tardive dyskinesia). Risk factors often prove more helpful as heuristic clues for researchers than as diagnostic alerts for clinicians. Evidence of causality is proportional to clinical importance of risk factors in screening.

The authors’ study is unusual with respect to the ubiquity of involvement of its sponsor (Neurocrine, Inc., the manufacturer of valbenazine (Ingrezza), approved in 2017 for treatment of tardive dyskinesia.

Disclosure accompanying the published study indicates that no one not an employee of, contractor hired by, consultant to, or speaker for Neurocrine or Teva (the manufacturer of deutetrabenazine (Austedo), was involved in its conception or execution. By its own admission, the sponsor designed the study, hired a contractor to implement a controversial methodology for developing consensus, and then hired additional contractors (one to collect, collate, and analyze the data, and interpret its significance; and the other, to write up the study for publication). The contributions of its authors are not attributed. Opportunity for sponsorship bias therefore extended from start to finish, and motivation for it to have been brought to bear acquires relevance  from the premise of the study (which was less of an investigation than a solicitation of opinion of participants, pursuant to a methodology fraught with a history of recurrent, uncorrected deficiencies (of arbitrary, non-random selection of topics and bibliographic content informing them, arbitrary criteria defining expertise required for inclusion in sample, vagueness of terminology fostering inequivalence of stated goals and conclusions from results, and violations of anonymity), adding substance to an impression of “loose tolerances,” better suited to evoking opinions than reconciling differences.).

The authors do not discuss why the Delphi methodology was chosen for a purportedly scientific study with a pretension of defining a standard of care for a subspecialty of medicine. Were the 25,000 psychiatrists in the U.S. to be informed by the study and prescribe the sponsor’s product (one of three pharmacotherapies in the authors’ consensus statement) intended for 500,000 persons in the U.S. who have tardive dyskinesia, the yearly cost would be approximately $40,000 each (not including the once per lifetime laboratory determination of the CYP2D6 genotype)

Clinically significant errors in this study’s recommendations derive from misunderstanding the literature in one case and obliviousness of it in the others, while the omission of a rationale for censoring the literature on tetrabenazine at the beginning of the study’s “methods” section trips a wire suggesting sponsorial intervention.

Tetrabenazine has a history, and most of it was excluded from the core bibliography supplied to the participants before the beginning of the consensus development process.

  Fewer than half of valbenazine-treated patients have shown clinically significant improvement of tardive dyskinesia, The results of Neurocrine’s phase 3 “Kinect” study of valbenazine were consistent with those of tetrabenazine’s sixty-year clinical history apropos of tardive dyskinesia: in Neurocrine’s Kinect study, forty percent of patients with tardive dyskinesia who were treated with valbenazine showed fifty-percent or greater reductions of tardive dyskinesia ratings. (In a randomized, controlled study some years ago, ginkgo biloba did about as well; and so, in another randomized, controlled study, did Vitamin E in a subgroup of younger persons with tardive dyskinesia whose symptoms were relatively mild).  

Inclusion of evidence published earlier than 2007 also allow more accurate and realistic assessment of tetrabenazine’s risk of adversity than did that derivable exclusively from the post-2007 literature used as “talking points” in the authors’ study. Few of this era’s prescribers have experience with valbenazine (designated a VMAT2 inhibitor, acronymic for vesicular monoamine transporter protein 2, which, when inhibited prevents uptake and sequestration of dopamine, other catecholamines, and indoleamines, such as serotonin, within intracellular storage vesicles, where they are protected from metabolism by monoamine oxidase). VMAT2 inhibitors deplete dopamine for synaptic release, reduce synaptic dopamine concentrations, and decrease the likelihood of interactivity of dopamine with its postsynaptic receptor and thereby reduce or interrupt dopamine signaling. Tardive dyskinesia is purported to be the macroscopic manifestation of excessive postsynaptic dopamine signaling, resulting from proliferation of dopamine receptors in compensatory response to chronic reduction of dopamine-receptor interactivity due to neuroleptic-induced dopamine receptor antagonism. Whether valbenazine mitigates tardive dyskinesia by dint of that hypothetical mechanism of action is not certain. Better established are the actuality that VMAT2 inhibitors eventuate in depletion of dopamine in the basal ganglia; that that depletion of dopamine is associated with Parkinson’s disease by dint of degenerative change of the substantia nigra; that depletion of dopamine results in a functional equivalent of neuroleptic-induced reduction of dopamine signaling due to dopamine receptor antagonism; that VMAT2 inhibition by tetrabenazine, valbenazine, and deutetrabenazine, like neuroleptics, induce akathisia, an extrapyramidal adverse effect; that VMAT2 inhibitors have been used by basic researchers to induce dopamine depletion and nigral degeneration to create a model of human parkinsonism in mice, and that parkinsonism is an established adverse effect of tetrabenazine in up to 30 percent of treated patients.

Tetrabenazine and derivatives additionally induce and exacerbate clinically significant depression in approximately 20 percent of treated patients. (Depression and parkinsonism are the second and third most common adverse effects after somnolence.) More speculative but compelling in a context where its occurrence has been demonstrated in diverse cell types is the added risk of oxidative stress from intracellular accumulation of products of oxidative metabolism of dopamine attributable to its desequestration by VMAT2 inhibitors. Oxidative stress, like dopamine receptor supersensitivity, has been invoked as a putative etiology of tardive dyskinesia and degenerative diseases of the central nervous system, generally.

Tetrabenazine is not a highly effective drug for tardive dyskinesia, nor are its significant adverse effects rare. It is no more a “turn-key” pharmaceutical than is reserpine, which was documented to induce clinically significant depression during its use as an antihypertensive, and to induce clinically significant depletion of dopamine, norepinephrine and indoleamines, such as serotonin. The search for a better tolerated substitute for reserpine led to the synthesis of dozens of benzoquinoline derivatives. Tetrabenazine was born of that search in the late 1950’s and is chemically, pharmacologically and metabolically nearly identical to valbenazine and deutetrabenazine. Tetrabenazine, like the latter, is prescribed for hyperkinetic movement disorders, including tardive dyskinesia, and has been since the 1960’s. It has earned mixed grades for efficacy and tolerability since, and whatever advantage once accrued to its use by dint of cost has long been exploded by increments that now sum to a yearly cost similar to that of its proprietary derivatives valbenazine and deutetrabenazine.


   The authors acknowledge that optimal treatment will for most patients include continuing neuroleptic exposure, and include in their recommendations substitution of less potently risk-conferring second- and subsequent-generation neuroleptics at lowest effective dosages. Their recommendation for substitution is clozapine.

Yet, neither evidence nor the logic of benefit and risk support the authors’ recommendation of clozapine as first-line treatment, given the distinctive risk of clozapine-induced agranulocytosis (with a yearly incidence of 0.4 percent) and granulocytopenia (yearly incidence, 3.0 percent), which prompts more frequent monitoring for agranulocytosis after its detection, along with evidence identifying clozapine with the highest propensity among the atypical neuroleptics to induce weight gain, other metabolic adversities, seizures, dysphagia, bowel obstruction, urinary retention, and central anticholinergic syndrome.

Moreover, according to recent findings of comparative risk of tardive dyskinesia among atypical neuroleptics, the yearly incidence of tardive dyskinesia induced by clozapine (0.24) is comparable to those of olanzapine (0.25) and quetiapine (0.38) and six-fold higher than that of aripiprazole (0.045). Evidence therefore recommends aripiprazole rather than clozapine as the alternative to be preferred.  Given lower propensities than those of other the latter-generation neuroleptics to induce hyperprolactinemia and endocrine-associated complications, metabolic adversity, extrapyramidal adverse effects, complications of central and peripheral anticholinergic activity, and its low likelihood of QTc prolongation.

A recent systematic review and meta-analysis of 15,000 patients treated with atypical neuroleptics for long durations found clozapine had a yearly incidence of tardive dyskinesia (0.24) similar to that of olanzapine (0,22) and quetiapine (0.38) and about six times higher than that of aripiprazole (0.045). Consideration of other adversities of clozapine, such as agranulocytosis, metabolic adversity, and complications of antimuscarinic activity (such as bowel obstruction, urinary retention, dysphagia, and central anticholinergic toxicity) indicates that clozapine’s benefit to risk ratio is not the best among the atypical antipsychotics. (At this time, the best ratio is that of aripiprazole, which, in addition to the lowest yearly incidence of tardive dyskinesia, rarely induces hyperprolactinemia (with gynecomastia) or other complications of hormonal dysregulation in women, and potentially increased risk for developing some types of breast cancer; uncommon weight gain in adults and other metabolic adversities, very low rates of sedation or fatigue, rare QTc prolongation, and low likelihood of extrapyramidal symptoms). Effective for both primary psychotic symptoms of schizophrenia and for acute mania, aripiprazole and clozapine have similar spectrums of therapeutic effectivness; aripiprazole’s propensities to induce neuroleptic adverse effects are lower than those of clozapine (the exception is akathisia, which has been reported in a higher proportion of aripiprazole- than clozapine-treated patients. Moreover, yearly incidence of tardive dyskinesia in aripiprazole-treated patients (0.045) was distinctively lower than those of other second-generation neuroleptics. Olanzapine, like clozapine, and to a lesser extent quetiapine are ssociated with clinically significant weight gain, abnormal glucose tolerance, and hyperlipidemia (increases of triglycerides, total serum coholesterol and low-density lipoproteins).

Compared with risperidone (reported significantly more likely to induce acute extrapyramidal adverse effects than are other atypical neuroleptics (compared in aggregate and individually), and among the most potent of both first- and second-generation neuroleptics to induce hyperprolactinemia and its clinical complications associated therewith (gynecomastia in males, menstrual abnormalities in females of reproductive age, sexual dysfunction in males, along with plausible speculation of higher risk of developing some types of breast malignancies (presently uncorroborated), significantly lower yearly incidences of tardive dyskinesia have been reported in patients treated chronically with aripiprazole, olanzapine, quetiapine, and clozapine, among which that of aripiprazole (0.045) was one-fifth to one-sixth those of the other three (which ranged from 0.22 to 0.38). If preliminary data reporting lowest yearly incidences of tardive dyskinesia of aripiprazole, clozapine, olanzapine, and quetiapine and further corroborate separation of aripiprazole from clozapine, olanzapine, and quetiapine with respect to agranulocytosis, metabolic adversity, and hyperprolactinemia, the most appropriate recommendation for an alternative neuroleptic in persons with tardive dyskinesia who need ongoing neuroleptic treatment is aripiprazole. Moreover, the least appropriate alternative is risperidone, with its separable propensity to induce acute extrapyramidal adverse effects and hyperprolactinemia

        Another of the study’s recommendations that warrants correction devolves not from misinterpretation and/or obliviousness of evidence, but from a lack of evidence. The study’s admonition about anticholinergic agents, which, it implies, citing no investigative study or observational report, and instead giving as reference hearsay and generalization in informal clinical reviews or speculation about inverse correlation of “cholinergic-dopaminergic” neurotransmission, as evidential basis that anticholinergic agents increase risk of and/or exacerbate tardive dyskinesia. (In fact, the authors do not make explicit why they favor dosage reduction or discontinuation of anticholinergic agents, but rather imply that the latter is what they recommend–no reason given and only the previously mentioned hearsay and generalization referenced. Neither risk of developing nor symptomatic worsening of tardive dyskinesia has been substantiated over a long history of investigation or opportunity for spontaneous reportage, which has produced only opinion marinated in confirmation bias and mistaking speculation for evidence. In fact, included in actual systematic or observational evidence that qualifies as evidence is a report of improvement when the anticholinergic biperiden was added to the polytherpeutic regiments of two patients, and in two others, no observable improvement or worsening after addition of biperiden.

As compelling as empirical observation might be, it is clinical experience that compels recommendation about anticholinergic agents: patients who have tardive dyskinesia and need ongoing neuroleptic treatment may develop extrapyramidal adverse effects that anticholinergic agents may mitigate. Neuroleptic-induced parkinsonism and dystonia are evidence-based treatments for both; still unresolved is whether or to what extent acute neuroleptic-induced akathisia may be. Many anecdotal reports describe absence or insufficiency of anticholinergic agents when given for akathisia at dosages typically effective for parkinsonism or dystonia, and in that context, robust effects have been described when beta adrenergic blockers and/or benzodiazepines have been resorted to. In persons who develop acute, neuroleptic-induced parkinsonism or dystonia, anticholinergic agents (such as benztropine or trihexyphenidyl) are effective and safe. Yet, because adverse effects of anticholinergic agents are dose-dependent and may be serious, dosing should be conservative and alternatives sought when conventional dosages are insufficiently effective. “Heroic” dosing may result in central anticholinergic syndrome, ileus with bowel obstruction, dysphagia, or urinary retention. Not acceptable is avoidance of approved dosages of anticholinergic agents in persons who develop acute neuroleptic-induced parkinsonism or dystonia; should akathisia prove insufficiently responsive to conventional dosages of anticholinergic agents, coadministration or replacement of the latter with a beta adrenergic antagonist (such as propranolol at a low starting acute “test” dosage of 20 or 40 milligrams (less if patients are taking potent inhibitors of CYP2D6, such as bupropion, paroxetine, or fluoxetine or are known to be CYP2D6 “poor metabolizers” by virtue of alleles associated with low activity)–and/or low initial oral dosage of a benzodiazepine (such as clonazepam (0.5 to 1.0 milligram), diazepam 5 to 10 mg, or lorazepam 0.5 to 1 mg orally, with response-based adjustment for dose and interdose interval). Because principal hepatic metabolic enzymes differ for beta adrenergic antagonists (CYP2D6) and benzodiazepines (CYP3A4, CYP2C9, coadminisration may allow therapeutic drug-drug pharmacodynamic interactivity in which an increment of effectiveness can be achieved without increasing the bioavailability of either over what it might be during monotherapy with either.

   Not elaborated beyond mention of cumulative exposure as dosage, risk for developing tardive dyskinesia or other manifestation of drug exposure contributed by pharmacokinetic drug-drug interactivity is conventionally expressed by blood drug or drug metabolite concentrations per unit dose in fluid compartments in equilibrium with brain tissue concentrations. Preponderantly dependent during clearance upon hepatic metabolism by cytochrome enzymes CYPs 1A2, 2D6, and 3A4, of which they are substrates, neuroleptics may achieve higher bioavailabilities during coadministration of potent inhibitors of the latter. Decrements of comparable magnitude have been reported during coadministration of hepatic inducers of CYPs 1A2 and 3A4, and have potential to develop in CYP2D6 ultra-rapid metabolizers (whose incremental enzymatic activity devolves from an extra wild-type allele). Severity of clinical adversity is proportional to dose (exposure) within typical dosage ranges; it may accompany hepatic enzyme inhibition with increments of parent neuroleptic or hepatic-induction-related increments of neuroleptic metabolite concentrations by dint of their own activity or in consequence of back-conversion to parent drug (as reported in association with clozapine-N-oxide, which has an affinity for dopamine receptor sites one fourth that of clozapine and withal is back-convertible to clozapine).

   So, too, warranting mention is critical dependence of tetrabenazine upon CYP2D6 for hepatic metabolism during clearance, with exiguous evidence so far negative for clinically significant pharmacokinetic drug-drug interactivity with inhibitors of CYP2D6 in tetrabenazine-treated humans with hyperkinetic movement disorders; the single adequately powered investigation that has so far looked for differences in tetrabenazine bioavailabilitin association with genotypes conferring poor, normal, extensive, and ultra-rapid metabolic activity failed to find them. The reason for contraindication of monoamine oxidase inhibitor coadministration with tetrabenazine is risk of increased extravesicular concentrations of catecholamine or indoleamines, which increase when those amines cannot be sequestered in intracellular vesicles (and thereby protected from metabolism by extravesicular monoamine oxidase) because tetrabenazine inhibits their vesicular uptake. That same inhibition may increase concentrations of toxic oxidized metabolites of dopamine—a scenario speculated by some to contribute to tetrabenazine adverse effects, particularly parkinsonism. The most potent inhibitors of CYP2D6 among the psychiatric pharmaceuticals are bupropion, paroxetine, and fluoxetine, which have plausible potential to raise tetrabenazine bioavailability by clinically significant increments and conduce to tetrabenazine-induced adverse effects.

   The Neurocrine-sponsored study by Caroff, Citrome and colleagues herein reviewed (ibid.) cannot be regarded as having offered optimal recommendations for inclusion in a consensus statement, given its excessive and continuous opportunity for exercise of bias (and failure to explain why pre-2007 literature, containing a preponderance of evidence pertaining to clinical experience with tetrabenazine was witheld from the bibliographic core supplied to participants); its preference for clozapine as a best alternative neuroleptic for persons with tardive dyskinesia who require ongoing treatment; its endorsement of its product, valbenazine, and other tetrabenazine derivatives as first-line treatment when no evidence suggests advantage of coadministering it with a neuroleptic, relative to neuroleptic monotherapy, and its implied, erroneous admonition to avoid exposure to anticholinergic agents.

The authors’ recommendation to suspect tardive dyskinesia in persons with any abnormal movement of any one body part is ill-advised because of features ratable on the AIMS scale that are nonspecific (such as frowning or grimacing or nonspecific tremor) and may promote false positivity and diminution of antipsychotic effects after inappropriate dosage reduction. from reduced dosage treatment.

An issue of sponsorship bias in the study is suggested for reasons subsumed under situational cogency, unexplained censorship of evidence pertaining to tetrabenazine published before 2007 (when a preponderance of literature concerning tetrabenazine was published), and disregard for usual and customary methodologic safeguards for exclusion of bias and artifactivity.

Finally, an explanation has not yet been communicated by the Journal of Clinical Psychiatry for providing a platform and vehicle for dissemination of a seriously flawed study inappropriately published as a consensus statement for prescribers. Critical issues herein broached and discussed were not considered sufficient cause for demur, resulting in no redaction of the study’s content or its sponsor’s failure to explain critical omissions that may represent censorship of core content for participants. The Journal failed to honor a fiduciary obligation to its readership and owes it an explanation, an apology, or both.


Leave a Reply

Your email address will not be published. Required fields are marked *