Is Clozapine The Most Dangerous Drug?

 

The Times came out with an article last week that did not get enough commentary.  In my opinion it was sensationalized and that was evident in both the title Britain’s most dangerous prescription drug — linked to 400 deaths a year and subtitle Clozapine has transformed the lives of thousands of schizophrenia patients but its dangers are not understood, say the families of those who have died from it(1).

A good starting point is my experience with clozapine.  When I was a research fellow in 1985, I was interested in prescribing it for people with treatment resistant schizophrenia.  Those were the days before atypical antipsychotics.  The first atypical was risperidone and that was not approved until 1993. I applied for compassionate use of the medication to the FDA, but I was eventually called by the company who manufactured it at the time.  They told me that they had no intention of allowing me to prescribe the medication before it was released to the public. That was eventually done in 1989, but it was under very tight regulations. A serious and potentially fatal adverse drug effect was agranulocytosis and that caused a number of related deaths in Finland. That meant every prescription was on a week-to-week basis contingent on getting a CBC with differential count. There were parameters to hold or discontinue the medication based on the ANC or absolute neutrophil count. There were also several other serious side effects like excessive fatigue, somnolence, significant weight gain, metabolic syndrome, diabetes mellitus Type 2, sialorrhea, severe constipation that could lead to bowel obstruction, hypotension, tachycardia, and myocarditis that required close follow up.

The initial expense led to tight regulation of the drug at the state level because a significant number of patients were disabled and on public assistance.  For years I had to complete a form stating that the patient had schizophrenia, had been tried on other medications, and needed clozapine. Even then it had to be approved by a clinical pharmacist who was the head of the state program. Eventually as the medication cost decreased specific retail and institutional pharmacies took over and were focused primarily on coordinating the blood draws and week to week prescriptions. A generic form of clozapine was released in 1999, but in a randomized study of changing to the generic – outcomes were worse (2).

In addition to treatment resistant schizophrenia, movement disorders could be treated by changing the antipsychotic medication to clozapine. In those early days of treatment with only typical antipsychotics tardive syndromes like tardive dyskinesia, tardive akathisia and tardive Parkinson’s were apparent.  Other refractory syndromes like tremors, torticollis, and dystonias also occurred in routine clinical practice. The patient population I was treating at the time often experienced severe psychosis and movement disorders at the same and had found no effective treatment. It is difficult to explain how disruptive severe hallucinations and delusions can be. Many of these patients required total care and could not function independently. It was clear that they were suffering and distressed. Clozapine often provided the first relief they experienced in years.

The combination of severe psychosis and the need for close monitoring was not an easy task for the physician. The medical complications needed to be avoided, but many of them depended on patient self-report and even then, a high index of suspicion by the physician. A good example is clozapine induced myocarditis.  The typical early symptoms including tachycardia, shortness of breath, and chest pain are commonly reported in a patient population that includes people who are heavy smokers, overweight or obese, and may have tachycardia as a drug side effect rather than myocarditis.

The Times article looks at all deaths of people taking clozapine as well as specific complaints to the regulatory agency and concludes that 400 people die per year (7,000 deaths since 1990 when it was licensed for use).  There are an additional 2,400 reports of severe side effects to the Medicines and Healthcare predicts Regulatory agency (MHRA) per year. The following paragraph is the only qualifier:

“The figures are not conclusive proof that clozapine is the cause of death because they record deaths of people on the drug, not simply because of it. Those people are already seriously ill and at risk.”

The current overall death rate in the UK is 337/100,000. The article states there are 37,000 patients in the UK taking clozapine.  The expected all cause death rate in the clozapine cohort would be about 125 per year.  We know from international studies that the life expectancy of patients with schizophrenia is about 25 years shorter than the adult cohort.  With a median standardized mortality ratio (SMR) in schizophrenia of 2.58 (3) the expected death rate in this population would be 325 per year – but with the ranges noted in this review it could significantly higher. The limitation with the Times estimate is that all-cause mortality is not noted in the article since the assumption is that all the mortality is clozapine related.  

Are there more likely direct cardiovascular causes of death? Newcomer and Hennekens (4) pointed out the association between severe mental illnesses and cardiovascular disease and potential modifying factors including cigarette smoking, decreased likelihood of medical treatment for modifiable risk factors including undiagnosed diabetes mellitus, and decreased likelihood of acute care for cardiac events. They also cite the lack of coordination of care among clinicians who are treating cardiovascular morbidity and psychiatric clinicians.

It would be useful to know if regulatory agencies had clear thresholds for recalling dangerous drugs. The reality is far from ideal.  For example the FDA recalled heparin after 4 deaths and 350 adverse events, but in the case of rofecoxib it missed the fact that is may have caused 88,000 to 138,000 heart attacks and strokes.  In the case of rofecoxib the company ended up voluntarily recalling the drug.  That extreme range of complications suggests that pharmacovigilance may only be a partial solution – but a lot depends on getting clear data and doing the correct analysis.  In pharmacology there is a concept called the therapeutic index (see the supplementary below) defined as the difference between the therapeutic range and toxic range for a particular medication. That range can be specified as the dose or plasma level.  One limitation of that approach is that it lumps broadly toxic medications with those that only affect a few individuals.  See the paragraph below for further discussion.  It is difficult to find a measure that applies at both the individual and population wide level. 

The remainder of the Times article focuses on the impressions of the relatives of deceased patients and a series of “more clozapine cases” from a preventable death registry. The relatives are understandably upset by the death of their family members and point out that they noticed problems for some time and in one case felt that clozapine was forced on them.  In the case reports/brief vignettes – it is not clear if clozapine was the cause of death or not.  The interaction between cigarette smoking and clozapine plasma levels was included and this is very useful information for the public.  In the case reports – coroner findings rather than autopsy results were reported.

I did not have any success in locating the information that the Times had access to at the MHRA web site, but I am familiar with previous pharmacovigilance research in the UK.  That study (5) reviewed 526,186 medication incident reports over a 5-year period from 2005 to 2010.  Seventy five percent of the reports were from acute care hospitals and the remainder from primary care clinics. There were 271 deaths and 551 incidents with severe outcomes.  The top 5 medications in terms of deaths were (in descending order) opioids, antibiotics, warfarin, low molecular weight heparin, and insulin.  The psychiatric medication on the list included benzodiazepines (15 deaths) and antipsychotics (2 deaths) accounting for 3.28% and 0.85% of the combined death and severe outcomes. I do not have access to the clozapine prescriptions per year or any updated pharmacovigilance data from the NRLS system.  It seems likely if clozapine was really causing hundreds of deaths in the UK someone would have flagged this and had the drug pulled off the market.

Apart from the analytical flaws in this article what might be going on?  As I have written about many times on this blog – medical decision making both on the recommendation and acceptance side is probabilistic and there is a lot of subjectivity.  It can only be approached concretely as error or no error after decisions have been made and outcomes determined. Even the ideal informed consent does not assure anything near a good outcome. Physicians who have seen suboptimal or overtly problematic outcomes know this – but patients less so and are generally hopeful that the newest treatment has something more to offer than what they have been doing. The equivalent bias in physicians is deciding that you are using an evidence-based treatment that is the best and wanting to maintain your patient on it when they are getting minimal benefit, significant side effects, or both. These decisions are complicated in the case of severe mental illness because of cognitive effects of the illness and possibly the medication.  It requires collateral information from people who know the person well and then another discussion with the patient.

Everything suggested in the previous paragraph takes time and more specifically – time with the most experienced member of the team. If my name is on the prescriptions, I want to be the person having these discussions.  I want to make sure that the patient, their family, and caregivers all know that I will never hesitate to discontinue a medication if it is not clearly more helpful than detrimental to the patient.  I want to make sure that every person in the room knows that at the time of the original informed consent discussion and that they can call me at any time with concerns. I want to make sure that I have enough medical knowledge to have the low threshold for diagnosing rare but serious complications and know what to do about them as quickly as possible.

In terms of a system of care whether that is in the US or the UK, all of that can be operationalized and monitored prospectively as a quality assurance project.  Even at that level there is a tendency of clinical and regulatory systems to be excessively rigid.  There is really no substitute for high quality treatment adhering to this cooperative process with ample opportunity for the patient or their surrogates to provide feedback to the responsible psychiatric staff and make active corrections – up to and including discontinuing clozapine - a daily opportunity.

 

 

George Dawson, MD, DFAPA

 

Addendum:  I contacted a clinical pharmacist recently who I had worked with in the past.  I offered to work on a pharmacovigilance system for the healthcare system we used to work for. I think it is the best way to get answers to these questions about the complications of medications and the associated prescribing practices.  I offered to work for free.  So far no return call. 

Supplementary 1:  One of the classic measures of a medication that may confer higher risk is the therapeutic index.  Therapeutic index is defined as the range between a therapeutic effect and a toxic effect.  Toxicity in this case can mean severe side effects that may be irreversible including possible death. That range could be in dosage but more precisely measured as plasma concentration.  This database lists 254 narrow therapeutic range drugs.  Clozapine is not on the list but in terms of psychiatric medications lithium, some antipsychotics, some anticonvulsants, and tricyclic antidepressants are.  Inspecting the list shows immediate limitations.  The chemotherapeutic agents listed are clearly more toxic than most of the other medications.  Non-steroidal anti-inflammatory drugs or NSAIDs are not listed despite significant mortality and morbidity.  Acetaminophen is not listed despite it being a leading cause of hepatic toxicity, liver transplantation and overdose death.

From a personal standpoint - I currently take 2 of the drugs on this list and use acetaminophen exclusively for pain.

https://go.drugbank.com/categories/DBCAT003972

References:

 

1:  O’Neill S.  Britain’s most dangerous prescription drug — linked to 400 deaths a year.  The Times, Sunday January 14, 2024.

2:  Kluznik JC, Walbek NH, Farnsworth MG, Melstrom K. Clinical effects of a randomized switch of patients from clozaril to generic clozapine. J Clin Psychiatry. 2001;62 Suppl 5:14-7; discussion 23-4. PMID: 11305843.

3:  Bushe CJ, Taylor M, Haukka J. Mortality in schizophrenia: a measurable clinical endpoint. J Psychopharmacol. 2010 Nov;24(4 Suppl):17-25. doi: 10.1177/1359786810382468. PMID: 20923917; PMCID: PMC2951589.

4:  Newcomer JW, Hennekens CH. Severe mental illness and risk of cardiovascular disease. JAMA. 2007 Oct 17;298(15):1794-6. doi: 10.1001/jama.298.15.1794. PMID: 17940236.

5:  Cousins DH, Gerrett D, Warner B. A review of medication incidents reported to the National Reporting and Learning System in England and Wales over 6 years (2005-2010). Br J Clin Pharmacol. 2012 Oct;74(4):597-604. doi: 10.1111/j.1365-2125.2011.04166.x. PMID: 22188210; PMCID: PMC3477327.

6:  Alvir JM, Lieberman JA, Safferman AZ, Schwimmer JL, Schaaf JA. Clozapine-induced agranulocytosis. Incidence and risk factors in the United States. N Engl J Med. 1993 Jul 15;329(3):162-7. doi: 10.1056/NEJM199307153290303. PMID: 8515788.

7:  La Grenade L, Graham D, Trontell A. Myocarditis and cardiomyopathy associated with clozapine use in the United States. N Engl J Med. 2001 Jul 19;345(3):224-5. doi: 10.1056/NEJM200107193450317. PMID: 11463031.

8:  Siskind D, Sidhu A, Cross J, Chua YT, Myles N, Cohen D, Kisely S. Systematic review and meta-analysis of rates of clozapine-associated myocarditis and cardiomyopathy. Aust N Z J Psychiatry. 2020 May;54(5):467-481. doi: 10.1177/0004867419898760. Epub 2020 Jan 20. PMID: 31957459.

9:  Medicines and Healthcare products Regulatory Agency (MHRA) Drug Safety alerts issued on clozapine  https://www.gov.uk/drug-safety-update?keywords=clozapine  Previous alerts issued on the risk and dangers of smoking cessation, metabolic syndrome and weight gain, therapeutic drug monitoring, intestinal obstruction, and drug interactions. All published 2020 or earlier.

 

Photo Credit:

Eduardo Colon, MD - much appreciated. 

COVID-19 No "Worse" than The Flu?

It is very common these days for people with varying motivations to make the argument that that COVID-19 the current pandemic caused by the SARS-CoV-2 virus is no worse than seasonal influenza.  After presenting that premise, the conclusions are typically that there is really no reason to implement social distancing, stay at home orders, and all of the additional precautions currently in place to prevent the spread of the virus.

The important qualifier here is the need to include how deaths from both illnesses are estimated.  The CDC is very clear that it does not know the exact number of influenza deaths each year but it estimates them from statistical models. This is nothing new and they have been using this procedure for decades.  The reasons include the fact that influenza deaths are not reportable at a national level, although pediatric influenza deaths are.  It is also not possible to know if influenza is the proximate cause of death because the death may occur weeks later as a result of a secondary infection or an exacerbation of a chronic medical condition by the influenza infection.  In these cases, influenza may not be listed as a secondary infection.  Finally - not everyone who dies from an influenza-like illness (ILI) has influenza and not everyone who dies from ILI is tested for influenza. The influenza death estimates are not based on death certificates for that reason.  

There is considerable variability in mortality estimates based on the model being used.  A description of their current methodology and its limitations is available at this link.   There is a similar limitation of COVID-19 related deaths and the CDC has a specific reporting procedure suggested for that process.  CDC clinical criteria and lab testing is further specified to determine if COVID-19 is an underlying cause of death. COVID-19 can be reported as "probable" or "presumed" based on clinical judgment.  Three examples are given in the linked document in how to fill out the death certificate. The main difference is that COVID-19 mortality depends on deaths certificates and influenza mortality does not.  At least for now.

Looking at the CDC death rate estimates for influenza over time looks like this (click to enlarge any graphic):

But looking at the raw data based on death certificates looks like this:

Looking at the typical influenza season going from 2019 (week 43) to 2020 (week 15) shows that the raw death certificate data for pneumonia is 90,369 and for influenza it is 7,591.   


The argument typically is made looking at disease mortality and the raw number of deaths are not used. The CDC and other agencies report rates per 100,000 to correct for differences in population.  The current mortality rate for COVID-19 as of today is shown below - even though the lead graphic illustrates that deaths are continuing to increase at this point.:


COVID -19 Deaths  CDC Page

Total Deaths: 37,158

Death Rate/100,000: 11.3

First Case January 21,2020


Influenza Deaths

Total Deaths:  24,000 - 62,000  2020 estimate based on above data and methodology

Death Rate/100,000:  7.3 - 18.9

First Case October 2019


The second consideration is that the COVID-19 pandemic is clearly not over. Different geographic areas in the US are at different points in the curve that depicts new cases.  The key point on that curve is the inflection point where the new cases per day go from a linear increase to an exponential one.  A panel of 20 experts in infectious disease modeling was referenced as the source for a recent White House estimate of 240,000 deaths by the end of 2020 (8).  If you look at that reference there is a wide confidence interval.  That is four times the CDC estimate of 2020 influenza deaths.  For comparison there were 675,000 deaths in the USA in the 1918 pandemic.

How does this information assist with the analysis of rhetoric?

1.  COVID-19 is no worse than seasonal flu: 

First off, influenza is a severe infection and can't be trivialized. Everyone who is able to should get an influenza vaccination. Based on the available data - is certainly seems that COVID-19 is as bad and much worse in the worst case scenario.  It is at least on par with modern CDC influenza death rates estimates over the past 20 years and based on the current number of deaths is likely to exceed the 2020 estimate for influenza deaths.  The outliers for the White House estimate in the survey are much higher - in some cases exceeding the mortality from the 1918 pandemic.  These estimates are also based on current rates and if the country is "opened" and stay-at-home orders, wearing masks and social distancing guidelines are abandoned it is very likely that there will be secondary spikes and prolonged exposure to the virus. Does anyone really want to take the chance of this virus killing more people than the 1918 pandemic?  

2.  COVID-19 rates are inflated based on inaccurate reporting:

This meme was reinforced by a physician appearing on a conservative talk show who stated that he only reports what he considers to be the underlying cause of death on death certificates.  The example given was that if the patient died of pneumonia - the cause of death was pneumonia and would not speculate on what caused the pneumonia.  The implication being that death certificates are highly accurate and the suggested reporting guidelines for COVID-19 will result in over-reporting the condition.  The information clearly shows that this is not the case. The actual numbers of deaths due to influenza as reported on death certificates are a fraction of the number estimated by the CDC.  The CDC plainly states that they have no idea how many people actually die from influenza and for decades they have estimated the number based on hospitalizations, hospital mortality, and other surveillance numbers.  COVID-19 death counts are made on the basis of death certificates.  Even though death certificates are not perfect, it is likely that many more people who die from COVID-19 are tested than people who die from influenza. It will be interesting to see if the CDC develops statistical models for COVID-19 to see if the current deaths are an underestimate like influenza. 

That is my brief look at these two arguments that are being used to suggest that the current environmental approach to virus containment are unnecessary.  I am also reminded of an old statistical concept called face validity.  Briefly stated that would mean the null hypothesis of no difference in death rates makes sense.  Given that COVID-19 has risen to the highest daily cause of death in the US, that hospital and ICU beds in many locations are overwhelmed, that there is a catastrophe in New York City at this point, and it has led to the only mass shortage of personal protective equipment, ventilators, and medical gear that I can recall in my 3 decade career - I don't think that it does make sense.

And this is exactly not the time to trivialize this pandemic.


George Dawson, MD, DFAPA


References:

1: CDC:  Frequently Asked Questions about Estimated Flu Burden.  Link

2: CDC:  Estimating Influenza-Related Deaths.  Link

3: National Center for Health Statistics. Guidance for certifying deaths due to COVID–19. Hyattsville, MD. 2020. Link

4: Reed C, Chaves SS, Daily Kirley P, Emerson R, Aragon D, Hancock EB, et al. Estimating influenza disease burden from population-based surveillance data in the United States. PLoS One. 2015;10(3):e0118369

5: Rolfes, MA, Foppa, IM, Garg, S, et al. Annual estimates of the burden of seasonal influenza in the United States: A tool for strengthening influenza surveillance and preparedness. Influenza Other Respi Viruses. 2018; 12: 132– 137. https://doi.org/10.1111/irv.12486

6: Centers for Disease Control and Prevention. Estimated influenza illnesses and hospitalizations averted by influenza vaccination – United States, 2012-13 influenza season. MMWR Morb Mortal Wkly Rep. 2013 Dec 13;62(49):997-1000.

7: Reed C, Kim IK, Singleton JA, Chaves SS, Flannery B, Finelli L, et al. Estimated influenza illnesses and hospitalizations averted by vaccination–United States, 2013-14 influenza season. MMWR Morb Mortal Wkly Rep. 2014 Dec 12;63(49):1151-4.

8:  Thomas McAndrew Spring March 25, 2020 COVID19-Expert ForecastSurvey6-20200325.pdf 


Graphics credit:

Lead graphic is from Our World in Data licensed under Creative Commons BY-SA and may be freely used for any purpose. 

All others are from the CDC under public domain.

Relief For the Sleep Deprived?

Sleep is a major problem for the majority of people who I see in clinical practice.  It is both a diagnostic feature and a primary disorder.  It is not uncommon for me to see people in their 30s or 40s who have had consistent sleep problems since childhood.  Environmental, medical problems, and alcohol/substance use are also common causes of insomnia.  After cessation of opioids, cannabis, or alcohol there can be disrupted sleep that lasts for months or longer. The treatment of insomnia is partially effective.  Behavioral methods like sleep hygiene measures and CBTi are useful for some people.  Medications can be helpful but they are a mixed bag for practitioners.  Sleep medications that are typically recommended have significant side effects including tolerance to the sedative effects that can lead to dose escalation and addiction.  The non-FDA approved medications like trazodone are widely used but routinely criticized in the literature for not having enough of an evidence base.  Physicians often face patients who are not sleeping well and ask for practical ways on catching up.  The news media lately has a lot of stories about the dangers of sleep deprivation creating some desperation in the sleep deprived population.  A common question is: "Can a sleep deprived person make up for lost sleep?" 

There was a very interesting study released by a research group this month on sleep and whether or not the sleep deprived can make up for lost sleep on the weekends.  The study looked at 38,015 participants in the Swedish National March Study who returned a general health questionnaire on medical history and lifestyle in 1997.  There were two questions about sleep:

How many hours  approximately, do you sleep during a workday/weekday night?

How many hours approximately, do you sleep per night on days off? 

The authors considered short sleep < 5 hours per night and long sleep > 9 hours per night.  The considered days off to be the equivalent of weekend sleep and simplified the response categories to reduce cells with low numbers of subjects.  The reference category was considered to be 7 hours.  The formed the following 6 categories based on that sleep classification and the pattern over the weekday/weekend (S=short, M=medium, L=-long):  SS, MM, LL, SML, ML, and LS.  Patient were following to the endpoints of death, emigration or study termination on December 31, 2010.

The authors used a Cox proportional hazards model with attained age to estimate mortality hazard ratios and 95% confidence intervals for each group adjusted for a number of variables including sex, BMI, smoking status,  physical activity, alcohol intake, educational level, shift work, and a weighted index based on an inpatient register.

The main finding with the correlations of mortality with short weekend sleep.  For subjects less that the age of 65, short weekend sleep was associated with a hazard ratio (HR) or 1.52 95% CI 1.15-2.02.  In other words subjects with short weekend sleep had a 52% greater mortality rate.  There was no different in mortality for short weekend sleep in subjects older than 65 years of age.  Forest plots were provided to look at adjusted and unadjusted HR across 5 sleep categories (≤ 5 hrs, 6 hrs, 7 hours, 8 hrs, ≥ 9 hrs).  A weekend sleep duration of ≤ 5 hours in subjects less than 65 clearly had the highest mortality ratio. In other analyses short sleep on both the weekdays and weekends and consistently long sleep were also associated with higher mortality.

Interestingly from a psychiatric perspective self reported sleep medication use did not alter the outcomes.  Sleep medication use was reported in every sleep category by 9.5 to 28% of the subjects in those categories (the short sleepers reporting more medication use).  Snoring, napping, restorative sleep, general health and high work demand did not affect results.  The initial model also corrected for shift work.

This is very interesting research because it suggests that there is a way to catch up on sleep debt at least on a short term basis.  Chronic sleep debt like the kind that physicians endure in medical school and residency training is probably gone forever.  But in clinical practice, it is theoretically possible to sleep in on the weekends after getting 5 hour blocks during the week and erase that debt - at least from  mortality standpoint.  Even though the authors seem to be doing a lot of analysis from 12 data points on a survey - the  structure of that data allowed them to look at sleep from a different perspective than it is typically analyzed from.  In their introductory section, they discuss the typical analysis focuses on typical sleep patterns and there are no distinctions between weekday and weekend hours.  Analyzing that data typically results in a J-shaped mortality curve with the highest mortality for too little sleep or a U-shaped mortality curve with highest mortality for too little and too much sleep.

The authors discuss the strength of their study (large N, good follow-up) and the potential weaknesses (misinterpretation of the questions by some subjects). From their exclusion process they did a good job of cleaning up the sample.  Their recommendation for closer follow-up studies on a longitudinal basis with more frequent data points is a good one.  From a clinical perspective, it would be useful to know what the time frame is that would allow for the cancellation of sleep debt.  Does it all have to happen in the space of a week or can you sleep very long at the end of two or three weeks and get back on track?  There may be some insights from people with prolonged insomnia from substance use (cannabis, methamphetamine, opioids, alcohol) and how they recover.


George Dawson, MD, DFAPA

References:

1:  Åkerstedt T, Ghilotti F, Grotta A, Zhao H, Adami HO, Trolle-Lagerros Y, Bellocco R. Sleep duration and mortality - Does weekend sleep matter? J Sleep Res. 2018 May 22:e12712. doi: 10.1111/jsr.12712. [Epub ahead of print] PubMed PMID: 29790200.



Graphic:

Sleep duration on successive nights from the smartphone of a person who is off work on the 19th and the 24th and works 20-23 - showing total hours of sleep as 8.19, 5.15, 5.51, 5.45, 5.49, 8.17.  This is a workday/weekend pattern described by the authors in the study.