Updated: Nov 7
Are there potential problems in the latest study on antipsychotic medication reduction and discontinuation?
A study on antipsychotic medication reduction and discontinuation came out yesterday with fanfare. The fanfare was basically because the principal investigator is a self-proclaimed critical psychiatrist with many criticisms of psychiatric medication and the results of her trial contradicted the primary hypothesis of the study and that was:
“Our hypothesis was that antipsychotic reduction would improve social functioning with only a small increase in relapse rate.”
Relapse rate in this case was defined as rehospitalization and the authors subsequently state that they thought a 10% rate of relapse would be “acceptable.” The irony of this situation (ideology versus real world treatment) was not lost on anyone. Several people seemed to congratulate the authors on publishing results inconsistent with their ideology although the study was so embedded in the UK research infrastructure – I doubt that not publishing it would have been an option.
As a clinical trialist myself – the research seems to present several problems and creates several questions that could suggest that it was designed to optimize the likelihood that antipsychotic medication could be reduced and possibly discontinued. Before I get into those scenarios let me briefly summarize the results. The paper is open access and can be downloaded as well as another paper that describes the research protocol.
In the study there were two arms an antipsychotic maintenance arm (N=127) and a reduction arm (N= 126). Diagnoses were taken from clinical information and the clinical staff had treatment responsibility for the patients. In those patients who were randomized to dose reduction, a tapering protocol was suggested to the clinical staff and if it went well at some point the option for a more rapid taper or discontinuation was offered. The research staff monitored the protocol. Baseline and outcome measure included a number of checklists to assess side effects, sexual side effects, positive and negative symptoms, quality of life, and social outcomes at the reassessment points. Raters were blinded but the measures are essentially self report. The ultimate result was that the risk of adverse outcomes was worse in the reduction arm with no associated improvement in social functioning.
He are some potential issues that I noticed based on my experience in clinical trial design and on research review boards.
1. Recruitment – described in the following:
“Participants were recruited from 19 National Health Service Trust mental health organisations across England. Potential participants were identified initially by clinical staff or recruited through advertisements placed in clinical settings and social media; those patients who expressed an interest in participating were sent further information”.
Not enough information. What did the advertisements say? Were subjects aware of who was running this trial and what the goal of the research was? Were the patients asked why they were interested in participating in this trial? Were they asked what they think about taking a medication? Did the subjects have any exposure to the considerable press that the critical psychiatry group and the principal investigator generate? Descriptions in the lay press have been demonstrated to have significant effects on perceived side effects – even to the point of creating a nocebo effect (6) – is there any reason to think that a group emphasizing side effects and minimizing any therapeutic effects might have a similar impact? If that is the case – how would it affect this trial?
2. Inclusion/Exclusion criteria –
“Exclusion criteria included being considered by a clinician to pose a serious risk of harm to self or others were the individual to reduce their antipsychotic medication, being mandated to take antipsychotic medication under a section of the Mental Health Act, having been admitted to hospital or treated by a crisis service for a mental disorder within the last month, lacking capacity to consent, having insufficient spoken English, pregnancy, breastfeeding, and being involved in another trial of an investigational medical product; eligibility was assessed by researchers and confirmed by the Principal Investigator for the site.”
Practically all the exclusion criteria result in a population that may be more likely to discontinue antipsychotic medications with less difficulty. Consistent with this is the antipsychotic doses of both the reduction and maintenance arm of 300 mg chlorpromazine equivalents (on average). According to the Maudsley Prescribing Guidelines (4) 300 mg chlorpromazine is considered the minimally effective dose of medication for relapsing schizophrenia. Whether this was a representative sample of the 4109 patients put forward for research by clinicians a comparison of the demographics and medication doses would have been of interest.
Selection bias may also be evident in the Consort diagram (page 4). After subjects consented to be contacted by the research team (N= 958) – a total of 562 declined participation. Was that because they did not want to take the chance of randomization to a medication reduction?
3. Diagnoses – the diagnosis required was schizophrenia or non-affective psychoses with recurrent episodes. The diagnoses were taken from clinical records. Considerable heterogeneity is introduced with the non-specific category of psychoses with an unpredictable course for which the concept of maintenance medication was not intended.
4. The Dose Reduction -
The description of the dose reductions in the paper is confusing. It starts out describing individualized reductions every 2 months based on starting doses but at some point states the patient is allowed to discontinue the medication if the dose reduction has been going well or reduce at a rate of the equivalent of 2 mg haloperidol/day. 2 mg/day of haloperidol is not a slow reduction and it is a departure from reduction every 2 months. Some of the authors here have written about antipsychotic withdrawal reactions – how is the more rapid dose reduction or optional abrupt discontinuation justified?
4. Safety Monitoring/Informed Consent:
The more clinical trials I read (and I have read thousands) – the more I want to see the consent form that each patient signs. Some of the authors here continuously talk about medication side effects. In fact – the principle investigator (PI) has stated that in her opinion that modern psychiatric medications work in a "drug centered" rather than a disease centered model by producing side effects like sedation, cognitive impairment, dysphoria, and loss of libido (5). In that model, symptoms of mental illness are muted by side effects rather than effectively treated. The model essentially denies the possibility of effective treatment without medication side effects. Of course, there are medication side effects but consent forms also must contain a discussion of the risks of the intervention. How are they listed when the investigators do not believe they can be directly addressed? Were the subjects told about the risk from medication discontinuation of recurrent psychosis, suicidal thinking, and death? That seems especially relevant in a study where the intervention arm had twice as many deaths as the maintenance arm .
Along those same lines – the protocol paper for the study (2) states that a Data Safety and Monitoring Board (DSMB) assessed the ongoing safety of the protocol and made recommendation to a Programme Steering Committee providing independent oversight – even to the point of stopping the protocol if there was a substantial increase in adverse events related to the intervention. Was there a threshold? In this case why was that threshold not met? In the trials I have been involved with the PI and the physician responsible for monitoring safety (typically me) had to clearly delineate a safety plan if any of the research subjects developed medical or psychiatric complications from the intervention. In this case that responsibility seems to have been delegated to the clinicians originally treating the patient.
In the reported causes of death of the trial participants – how is the death of a research subject in the reduction arm attributed to antipsychotic medication when they have been on a low dose, were being followed clinically in an outpatient clinic, and their dose was presumably being reduced? One patient in each arm died of an “accidental overdose”. What medication was implicated in the accidental overdoses?
This protocol is also a case of shifting risk for the research to the clinicians. Here the research staff designs an intervention that likely will lead to worsening clinic status and the subjects are followed in a treatment as usual manner. Were any additional safeguards in place for that eventuality? For example – were the subjects informed that they could contact the principal investigator or research coordinator if things were not going well?
These all seem like significant safety questions to me.
5. Social Functioning Scale (SFS) to measure the primary outcome -
The measured results with this scale are in the top line of Table 2 at 6, 12, and 24 months. The scale has 79 items that are assigned to assess social functioning. Is there a problem with taking a cross sectional sample of people stabilized on medications and hypothesizing they will function better being tapered off antipsychotic medication? There is an obvious problem and that is there is no accounting for the improvement in social functioning due to the medication in the first place. In other words - what would the subjects have scored leading up to and during the episode of acute or recurrent psychosis - the reason they are taking the medication in the first place. What would the trajectory of these scores be over time? Stabilization of psychosis involves a lot more than treating hallucinations, delusions, and thought disorder symptoms. With stabilization there is an improvement in social behavior.
The design of the trial suggests that the problem began with medications rather than a significant psychiatric disorder.
There is a concept in clinical psychiatry and that is trying to get the patient as close as possible to their baseline level of functioning. That requires a knowledge of what they were like before the onset of illness and restoring as much functional and social capacity as possible. That also typically means minimal to no medication side effects if possible.
6. What is supported reduction of antipsychotic medication?
Is there a protocol that I missed? I could not find what this means anywhere in either the protocol or final paper or in the supplementaries. If I was tapering an antipsychotic medication I would meet more frequently with the patient, inform them of what we need to watch for, have additional caregiver and family involvement, and encourage them to call me at specific signs of early problems due to the dosage reduction. In a research protocol, research staff would call and check on how the subject was doing. I would call all of that treatment as usual (TAU) when it comes to antipsychotic medication reduction. Is supported reduction more than that? Even TAU has been implicated as a potential placebo enhancing effect. Did it have that effect on the intervention in this case?
7. The overstated conclusion:
“Our findings provide information for people with schizophrenia and related conditions about the probable medium-term impact of reducing the dose of their antipsychotic medication, and they highlight the need for collaborative decision-making based on the sharing and careful consideration of all the evidence.”
Actually, it doesn’t. This is what clinical psychiatrists do and more specifically it is what I did for 35 years of practice. I can still recall community psychiatry seminars with Len Stein, talking about dosage reductions of antipsychotic medications and the implication of a WHO international study looking at that problem in schizophrenia. That seminar was in 1986.
Collaborative decision making seems to be the latest term for informed consent and therapeutic alliance. Informed consent means that the patient is given enough information and discussion so that they can make a decision about the direction of their care including any medications, tests, or other interventions used. The therapeutic alliance is the affiliative relationship between the patient and physician aligned to address the patient's problems and diagnoses. It is by longstanding definition a collaboration.
What the authors did encounter but did not discuss was the tendency of people on antipsychotics to just discontinue them (several in the maintenance group did this), how much withdrawal was encountered, and why there were no group categorical differences in side effects with the taper. According to the Glasgow Antipsychotic Side-effect Scale (GASS) guidelines all subjects remained in the moderate side effect range. And if medications work through side effects as the critical psychiatrists say why did the subjects in the dose reduction group worsen?
Those are a few of the problems that jumped out at me as I read this paper and the associated backgrounder. As can be seen from the above discussion many of these design factors potentially optimize the intervention group in the direction of proving the authors’ hypothesis. It also limits generalizability to other clinical settings. That makes the result of the trial even more significant. It also raises some issues that seem more prominent in recent years as pharmaceutical conflict of interest seems to ring hollow. Is there an ideological conflict of interest and how is it determined? How does it affect research design, results, and the discussion of research findings?
The Recent Takedowns of Adult ADHD
Psychiatry seems doomed to argue endlessly about whether certain conditions exist or not and whether they can be characterized by written criteria. The latter condition is the most easily dismissed since clinical training is necessary to recognize conditions. You cannot just sit in an office, read the DSM and call yourself a psychiatrist. Whether conditions exist or not is more debatable but often slides into rhetoric that suggests inadequate training, ignorance, and/or significant conflict of influence or undue influence by the pharmaceutical industry. Consideration of the undue influence can easily be applied at the global level since Pharma has massive marketing efforts, direct to consumer advertising in the US, and at least one major political party pulling for them.
That brings me to the recent commentaries about adult ADHD (1, 2). The first reference (1) doubts that adult ADHD exists for the most part and sees the diagnosis primarily as the result of a marketing scheme by Eli Lilly for atomoxetine and ignoring affective temperaments and other states that may affect attention. Atomoxetine was invented as a norepinephrine reuptake inhibiting antidepressant and like other members of this class of drugs – it did not work for depression. Since it is not technically a stimulant it was tested for ADHD and found to be effective. It is unique relative to other ADHD medications and not surprisingly it was heavily marketed while on patent. The patent expired on May 2017. The years on the market patent protected were 2002-2017. The first references to the diagnosis of adult ADHD were noted in the 1980s. Reference 2 suggests that the diagnosis of ADHD in children in the US is around 2-3% with adult numbers half that based on the work of one author. Contrasting numbers of a lifetime prevalence in adults as 8.1% and surveys estimating current prevalence at 4.4% are described as “absurdly high” but qualified on methodology (surveys vs interviews). Some authors have the opinion that books published about adult ADHD like Ratey and Hollowell's Driven to Distraction were a major source of public interest in the diagnosis and instrumental in getting it into the public vernacular.
Before I get started – let me say that the only stake I have in this argument is making sure that the complexity of the situation is adequately described. Practically all the pro/con arguments in psychiatry are gross oversimplifications and based on what I know about the literature – I had no reason to expect that this was any different. I am already on record on this blog describing how to diagnose and treat ADHD and not fall into the common problems of misdiagnosis, prescribing to people with substance use problems, or prescribing to people who view these medications as performance enhancers. I have successfully treated adult ADHD with both on and off label medications and can attest to the fact that it is a valid and treatable diagnosis.
Let me start out by looking at the prevalence estimates. These figures are very popular in the press to indict diagnosticians in the United States compared with some European countries and sell more papers. The problem with prevalence estimate is that the range can vary significantly due to methodological differences in the surveys. That question was looked at (3) and the title of that paper asked if ADHD was “an American condition”. The authors reviewed 22 studies based on DSM-III criteria and 19 studies based on DSM-IV criteria.
Twenty prevalence estimates were done on the US and 30 were done in other countries.
They demonstrated that the range of prevalence across all studies was approximately the same and that ADHD was not just an American condition. Since then numerous prevalence studies have been done in other countries – more recently using DSM-5 criteria showing similar ranges.
On the issue of adult ADHD, a recent review looked at the issue adult ADHD and symptomatic adult ADHD prevalence by the 6 WHO regions (4). Their overall goal was to determine the worldwide prevalence of adult ADHD. They looked at the issue of persistent or childhood onset ADHD and symptomatic adult ADHD with no evidence of childhood onset and estimated the prevalence of those two groups separately. The pooled prevalence of persistent adult ADHD was 4.6% and for symptomatic ADHD it was 8.83%. These authors also looked at prevalence by a list of demographic factors, diagnostic criteria, addition to geographic areas as well as the decreasing prevalence by age groups.
Prevalence % (US vs Non-US) ranges or pooled
Faraone, et al (2002)
(9.1-12.1) vs. (5.8-11.2)
(7.1-12.8) vs. (3.9-10.9)
(11.4-16.1) vs. (2.4-19.8)
Polanczyk, et al
Pooled prevalence estimates of ADHD by geographic location.N= number of studies in each WHO designated location
North American (N=32) 6%
Europe (N=32) 4.5%
Oceana (N=6) 4.5%
South American (N=9) 12%
Asia (N=15) 4%
Africa (N=4) 8%
Middle East (N=4) 2.5%
Song, et al (2021)
Pooled estimates and ranges of Adult ADHD worldwide by WHO designated geographic areas
and ranges of Adult ADHD worldwide by WHO designated geographic areas
North America (N=3) 6.06%
Europe (N=10) 7.12%
Oceana (N=4) 9.67%
South America (N=3) 6.06%
Asia (N=1) 25.6%
Africa (N=1) 9.17%
Middle East (N=2) 16.58%
This study raises the issue of whether ADHD can be acquired rather than be a childhood onset illness. The reality is that there are many paths to acquired attentional deficit that have been treated over the course of my 35 years in the field. The best examples are neurodegenerative diseases, strokes, and brain injuries. Neuropsychiatrists have written about treating the associated cognitive, mood, and motivational deficits with stimulants. But a more relevant question is whether mechanisms exist that can result in people with none of these acquired brain injuries. The answer comes from modern genetics. Polygenic risk scores (of all diseases) suggest that there are high risk individuals who show no evidence of an illness as adults. These examples of incomplete penetrance are usually explained as environmental factors, additional genetic dynamics such as aging or protective factors. I see no reason why these factors could not occur in an ADHD genotype after childhood. The other significant genetic factor is spontaneous mutation or as a recent commentator put it: “You don’t die with the genome you were born with.” Psychiatry has focused on familial studies for the past 50 years, but it is likely that significant numbers of most conditions occur as the result of spontaneous mutations rather than strictly hereditary transmission. That is borne out in clinical practice every day.
The authors (1) make the argument that ADHD is not a “scientifically valid” diagnosis. They explain “these symptoms have not been shown to be the result of a scientifically valid disease (adult ADHD) and better explained by more classic and scientifically validated psychiatric conditions, namely diseases or abnormalities of mood, anxiety, or mood temperament.” Mood temperament is a stretch. It is rarely commented on in adult psychiatry and then in extreme cases. It is not contained in the DSM. Part of the reason is selection bias. Psychiatrists are seeing people who have failed multiple other treatments and I have referred to this as being the treatment provider of last resort.
Another factor is that ADHD is a quantitative rather than qualitative disorder – that is the cognitive symptoms are at the extreme end of normalcy and it is difficult to draw a line to demarcate illness from normal in many cases. A comparable example from medicine is hypertension. The cutoff for what is considered hypertension has varied significantly over the decades (9, 10) and even now considers antihypertensive side effects as a qualifier for treatment. That means that for any 2 people with the same marginally elevated blood pressure only one might get consistently treated. At one point hypertension was considered by some physicians to be a necessary compensatory mechanism that should not be treated (10). On the issue of quantitative aspects of psychiatric disorders in general – dimensional approaches are often suggested as a solution and the question is whether they work any better than the impairment criteria used in the DSM. That is especially true in a clinical setting where a patient is presenting with a clear problem that they are asking for help with
On the issue of validity, studies have been done demonstrating reliability and validity (8) on both the DSM criteria as well as various rating scales for adult ADHD that are consistent with the diagnosis. There have also been detailed discussions of how to approach the problem clinically (11). Those discussions include how to differentiate mood disorders from ADHD and how to approach the functional impairment criteria in the clinical interview.
That brings me to the issue of temperaments mentioned in reference 1. Temperaments have been researched in various contexts in psychiatry over the past decades. Most psychiatrists of my generation first heard about them on child psychiatry rotations and the work of Stella and Chess. In adults, temperaments are more descriptions of hyperthymia, cyclothymia, and dysthymia and are generally considered in the differential diagnosis of subclinical mood disorders. The best example is hyperthymia and it has been referred to both as a temperament and a personality. Hyperthymic people are generally high energy, require less sleep, and are social, talkative, and outgoing. They may be very productive and have increased libido relative to their peers. In clinical interviews they may say that their friends think they are “bipolar” and need to be treated. But careful interviewing demonstrates that they lack the symptom severity and degree of impairment necessary for a diagnosis of bipolar disorder. Ideally the initial interview results in that formulation and the psychiatrist can advise the person about why treatment is not necessary.
Reference 12 looks at the issue of temperaments in a retrospective controlled study of patients being treated with stimulants who were referred to a mood disorders clinic. The authors acknowledge the selection bias in their study design. I can not think of a better design to pick up misdiagnosed patients than this one. To cite one example – of the 87 amphetamine treated referrals only 50% had a past diagnosis of ADHD. The authors acknowledge that there is no standard way to determine affective temperaments and decide to use the TEMPS-A with a cutoff of 75% of the items. If you are able to find a copy of the TEMPS-A (it is not easy) – you will find a list of 50 true-false questions like “I’m usually in an upbeat or cheery mood.” The questions are reminiscent of the Minnesota Multiphasic Personality Inventory (MMPI) except there are far fewer questions. The scoring guide suggests that the TEMPS-A can discriminate between hyperthymic, cyclothymic, dysthymic, and irritable temperaments. It is validated in the usual ways. The relevant question is whether any diagnosis made with this checklist would deter you from treating a comorbid condition - like Adult ADHD? It is one thing to survey a misdiagnosed group with the TEMPS-A and consider the clinical implications, but another to consider the presenting problem possible ADHD and whether it should be treated.
The arguments in reference 2 about overdiagnosis, the existence of adult ADHD, and the idea that ADHD can occur in adults without a childhood diagnosis can be challenged with the facts and references provided here. The fact that we are in the midst of a multigenerational drug epidemic in an increasingly intoxicant permissive society does not mean that a diagnosis, treatment, or problem does not exist. It does mean that all psychiatrists from the moment they enter practice must exercise extreme caution when prescribing substances that reinforce their own use.
The most likely cause of overdiagnosis is not because adult ADHD does not exist, not because of drug promotion (most are generic including the non-stimulant alternatives), or because MDs are careless. There are basically two reasons. First – the difficulty of diagnosing quantitative conditions. Second – sociocultural factors that exist in the US. Performance enhancement is built on the myth that you can tune your brain (or any organ) with supplements, nutrients, or medications to become a superior human being. The reality is you can alter your conscious state to believe that – but in the case of stimulants it is unlikely. The only real performance enhancement occurs because you can stay awake longer to read more and there is some evidence that your belief system is altered so that you believe you are smarter (14). These are just two of the reinforcing properties of stimulants that can lead to accelerated use and addiction.
That is my brief summary of the complexity of this situation. For more on my approach to adult ADHD.
Heart Rate Variability
I have been following heart rate variability (HRV) on my watch and three different apps for the past several years. HRV is defined as the slight variations between R waves in the standard ECG recording. I have included an example below, illustrating the R-R’ intervals (or RRI) and how they might vary over time.
Since HRV became widely available as a measurement off a watch that is commonly worn by millions of people, the research on this measurement and the variable studied has increased significantly. For my purposes – HRV is thought to be an indicator of heart health and conditioning and possibly a marker of overtraining – but advice about that varies significantly. Some studies have shown that decreased HRV is associated with an increased risk of arrhythmias. My recent cardiac ablation and cardioversion seemed to present an ideal situation for further study.
Before getting into those details the physiology of HRV needs to be considered. The dominant heart rhythm of a normal heart is determined by the sinoatrial (SA) node. This node contains a population of spontaneously depolarizing cells that determine the rhythm and rate of the heartbeat. In addition to the neurophysiology of that cell population several additional factors affect both the rate and HRV. Primary among them is autonomic innervation from both the sympathetic and parasympathetic systems and their effect at the SA node. Parasympathetic fibers from the vagus nerve modulate slower firing through the neurotransmitter acetylcholine (ACh). Sympathetic fibers increase the rate of firing through the neurotransmitter norepinephrine (NE). NE has a longer half-life than ACh, but vagal tone is thought to be the most significant determinant of HRV. That is in line with several clinical observations including lower baseline heart rates in conditioned athletes and higher heart rates in people with less conditioning or in stressful situations.
What happened to my heart rate and HRV during the recent cardiac ablation for atrial fibrillation and subsequent cardioversion? To answer that question, I had to figure out how to get the data off my Apple Watch 5.0. The only approach I could find was to downloaded all of the collected Health App data as a CSV file and then plot it in Excel. There are some online sites that you can download the data to and then use the remote software for plotting, but I preferred to retain control over the data. If you decide to do that and have several years of data like I did – it takes a long time. It took about 5 hours in my case to download about 1G of data to a zip file. From there it is easy to open that file with Excel or other software and do the plots. A useful addition to the Health App would be able to download specific time intervals.
I have done 2 plots so far based on average daily HRV and hourly HRV as shown below.
The plots are interesting because it clearly shows an effect from the ablation, a 96-hour period of atrial fibrillation and atrial flutter, and the cardioversion. At the minimum the baseline HRV drops to a different baseline after the ablation. That is followed by a significant spike with the recurrence of afib/flutter. And then there is a return to the lower baseline after the electrical cardioversion. I rarely had any significant episodes over the course of a year and whenever I went back and reviewed HRV it was not significantly changed. Since all those episodes were typically less than 2 or 3 hours it may not have been long enough to see an HRV effect. Conversely spikes of 50-100 msec in the HRV recording were common and not associated with arrhythmias. In the case of the post ablation period the sustained rates were associated with spikes, but since atrial flutter is regular, the associated R-R’ intervals probably showed a more characteristic HRV.
I would expect to see an increase in vagal tone and therefore HRV just related to the sustained high rates over 4 days. If increased vagal tone correlates with increased HRV that does not seem to be the case in these graphs. The graphs also seem to indicate to me that there may be a structural element to HRV – either in the anatomical configuration of the conducting cells, their altered physiology, or a combination.
The main implication for me at this point is to cautiously restart my conditioning efforts and see what impact that has on the HRV baseline. A second question is whether my HRV will approach the pre-ablation baseline. Electrocardiograms (ECG) may provide some clues in that direction. I have listed them below for references. Significant changes occurred in the immediate post ablation ECG and the post cardioversion ECG.
An additional thought is whether non linear analysis of the RR intervals would yield more information and easily interpretable graphics. I have used some of these attractor plots in the past and also applied them to single electrode analyses of normal controls and patients with Alzheimer's disease. In terms of ECG analysis - see figure 5 in reference 2. In terms of theory - these attractor diagrams also imply changes in biological complexity at either the structural or functional level.
George Dawson, MD, DFAPA
ECG time course (1 -> 5 are in sequence):
1. Baseline - preop ECG
2. Post ablation ECG (following day):
3. Post ablation ECG - note anterior T wave changes thought to be consistent with procedure.
4. Precardioversion ECG showing atrial flutter at a high rate (day 5 of this arrhythmia; post op day 14).
5. Post cardioversion ECG showing NSR but flipped T waves in V1-V3.
6. ECG follow up 2 weeks after cardioversion showing T wave normalization in anterior leads.
Heart Rate Variability
Some recent recovery in HRV after a long period of low numbers in the 7-37 msec range following ablation and cardioversion.