Anti-psychotics, damn lies and statistics

Today, I was asked to clarify some information surrounding an earlier post about Carrie Fisher and the role that anti-psychotic medications may or may not have played in her death from a heart attack.  I appreciated the question which was about whether I’d seen the statistics put up by the Mad In America (MIA) blogger who wrote about Carrie Fisher (the blog which, incidentally, Dr Leaf had then uncritically decided to slyly try to regift it in the form of her newsletter article).

In the opening of her post, the MIA blogger said, “There’s an important question here. Is she one of the cases in point to explain why our community has a 25 year lower life expectancy?” and then threw in a table plucked out of context from a journal article.  At least, unlike Dr Leaf, the MIA blogger was intellectually honest enough to attach the source of the table, which was an article published in the European Heart Journal in 2012.

While the MIA blogger is certainly entitled to her opinion, I thought it was worth discussing the statistics in a bit more detail, if for nothing else than to give some context to the whole “anti-psychotics kill you” trope that keeps getting around.

First, there needs to be the proper context.  No one is denying that there’s a higher mortality rate amongst people with schizophrenia and other forms of psychosis, though I don’t see exactly where she got her “25 year lower life expectancy” line from. To me, that seems excessive.

Then to the study itself.  The paper that the table is extracted from is Honkola et al [1]. The study specifically examines the association between the use of different classes of psychiatric medications with the rate of sudden cardiac death during a coronary event (a heart attack, or angina).

In her post, the MIA blogger throws around a lot of numbers but she was loathe to put her numbers in the right context.  For example, she claimed that “smoking is four times safer than the older types of antipsychotics. And it’s twice as safe to smoke as it is to take any antipsychotic, including the newer ones”.  Except, her comparison is a fallacy of conflation – she’s comparing the all cause mortality of smoking (which is more like three-fold rather than two-fold, just FYI [2]) with the highly specific ‘sudden cardiac death during a heart attack’ mortality of the study she’s referencing.  It’s apples and oranges – the groups aren’t directly comparable.

Besides, even if her numbers were directly applicable, the positively immoral sounding four-fold increase in the rate of death sounds is just an association, not a cause.  There is a dictum in science, “Correlation is not the same as causation.”  Just because two things occur together does not mean that one causes the other.  There may be other explanations beside the medication that might explain that number, including but not limited to, statistical anomalies and lifestyle factors, and other factors not considered in the analysis.

There are other problems with relevance too.  Most of the numbers in the table were small and not statistically significant (that is, could have been related to chance alone).  The only strong numbers were for old anti-psychotics, phenothiazines, tricyclic antidepressants and butyrophenones, none of which are first line medications for psychosis or depression anymore.  Newer anti-depressants and the newer atypical anti-psychotics did not have a statistically significant association.

And, like I said before, this study is looking at the association between sudden cardiac death in people having a heart attack, which is a very specific form of mortality.  It’s not particularly applicable to everyone on the medications, so even if the 4- or 8-fold increase is rock solid, you can’t translate that statistic to everyone on anti-psychotic medications or anti-depressants, or Carrie Fisher for that matter since no one really knows how she died other than she had a heart attack.  The rest is just disrespectful speculation.

For me, rather than trying to take a table full of weak and inapplicable statistics and beat a conclusion out of them, a more useful thing would be to know the benefit or harm of anti-psychotics on all causes of death.  If anti-psychotics were really as poisonous as Dr Leaf and the MIA blogger portrayed, then all-cause mortality would be much higher in those exposed to the drugs versus those who were never exposed to the drug, which is why this study by Torniainen and colleagues [3] is particularly interesting, and in particular, this graph –

In this study, the chance of dying from any cause was significantly higher in those people with schizophrenia who were never treated with anti-psychotics compared to those who were treated.

Does this answer the question why there is a lower rate of mortality? Not really, because in fairness, this study also showed just an association between no anti-psychotics and a higher death rate.  It doesn’t specifically prove causation one way or another.

Does it show that we should throw anti-psychotics around like lollies, or that they are wonder drugs without any associated harm? No, they are medicines and need to be used responsibly.

It does show there’s a general benefit to anti-psychotics for people with schizophrenia so they’re not the toxic killers Dr Leaf and the MIA blogger try and make them out to be.

Anyone can cherry-pick weak statistics and bend them to suit their self-interested propaganda.  The remedy to damn lies and statistics is to look more broadly and consider the strength of the numbers and their context.  When we do that with the studies on anti-psychotic medications we see that they aren’t the evil killers that some people would like to make them out to be.

[1]        Honkola J, Hookana E, Malinen S, et al. Psychotropic medications and the risk of sudden cardiac death during an acute coronary event. Eur Heart J 2012 Mar;33(6):745-51
[2]        Jha P, Ramasundarahettige C, Landsman V, et al. 21st-century hazards of smoking and benefits of cessation in the United States. The New England journal of medicine 2013 Jan 24;368(4):341-50
[3]        Torniainen M, Mittendorfer-Rutz E, Tanskanen A, et al. Antipsychotic treatment and mortality in schizophrenia. Schizophrenia bulletin 2015 May;41(3):656-63

Should pregnant women still take antidepressants if they’re depressed? – SSRI’s and the risk of autism

As is my usual habit, I sat down tonight to do something useful and wound up flicking though Facebook instead.  Procrastination … avoidance behaviour … yeah, probably.  But at least this time it turned out to be rather useful procrastination, because I came across a science news story on Science Daily about a study linking the use of anti-depressants in pregnancy with an 87% increased risk of autism.

Actually, this is old news.  Other studies have linked the use of some anti-depressants with an increased risk of autism, such as Rai et al in 2013 [1].

The latest study to come out used data from a collaboration called the Quebec Pregnancy Cohort and studied 145,456 children between the time of their conception up to age ten.  In total, 1,045 children in that cohort were diagnosed with autism of some form, which sounds like a lot, but it was only 0.72%, which is actually lower than the currently accepted prevalence of autism in the community of 1%.

What the researchers got excited about was the risk of developing autism if the mother took an antidepressant medication at least at one time during her pregnancy.  Controlling for other variables like the age, wealth, and other health of the mothers, a woman who took an anti-depressant during pregnancy had a 1.87 times greater chance that her baby would end up with ASD, compared to women who did not take an anti-depressant [2].

An 87% increase sounds like an awful lot.  In fact, it sounds like another reason why anti-depressants should be condemned … right?

Well, like all medical research, you’ve got to consider it all in context.

First, you’ve always got to remember that correlation doesn’t always equal causation.  In this particular study, there was a large number of women being followed, and their children were followed for a long enough time to capture all of the likely diagnoses.  So that’s a strength.  They also tried to control for a large number of variable when calculating the risk of anti-depressants, which also adds more weight to the numbers.

Although the numbers are strong, studies like these can’t prove that one thing causes another, merely that they’re somehow linked.  It might be that taking anti-depressants causes the brain changes of autism in the foetus, but this sort of study can’t prove that.

Even if the relationship between anti-depressants and ASD was cause-and-effect, what’s the absolute risk?  Given the numbers in the study, probably pretty small.  With a generous assumption that ten percent of the study population was taking anti-depressants, the increase in the absolute risk of a women taking anti-depressants having a child with ASD is about 0.5%.  Or, there would be one extra case of autism for every 171 that took anti-depressants.

Hmmm … when you think of it that way, it doesn’t sound as bad.

You also have to consider the increase in risk to women and their offspring when they have depression that remains untreated, or in women that stop their anti-depressant medications.  There is some evidence that babies born to women with untreated depression are at risk of prematurity, low birth weight, and growth restriction in the womb, as well as higher impulsivity, poor social interaction, and behavioural, emotional and learning difficulties.  For the mother, pregnant women with depression are more at risk of developing postpartum depression and suicidality, as well as pregnancy complications such as preeclampsia, and an increase in high-risk health behaviour such as smoking, drug and alcohol abuse, and poor nutrition.  Women who discontinued their antidepressant therapy relapsed significantly more frequently compared with women who maintained their antidepressant use throughout pregnancy (five times the rate) [3].

So the take home messages:

  1. Yes, there’s good evidence that taking anti-depressants in pregnancy is linked to an increased risk of a child developing autism.
  2. But the overall risk is still small. There is one extra case of autism for every 171 women who take anti-depressants through their pregnancy.
  3. And this should always be balanced out by the risks to the mother and child by not adequately treating depression through pregnancy.
  4. If you are pregnant or you would like to become pregnant, and you are taking anti-depressants, do not stop them suddenly. Talk to your GP, OBGYN or psychiatrist and work out a plan that’s best for you and your baby.


[1]       Rai D, Lee BK, Dalman C, Golding J, Lewis G, Magnusson C. Parental depression, maternal antidepressant use during pregnancy, and risk of autism spectrum disorders: population based case-control study. Bmj 2013;346:f2059.
[2]       Boukhris T, Sheehy O, Mottron L, Bérard A. Antidepressant use during pregnancy and the risk of autism spectrum disorder in children. JAMA Pediatrics 2015:1-8.
[3]       Chan J, Natekar A, Einarson A, Koren G. Risks of untreated depression in pregnancy. Can Fam Physician 2014 Mar;60(3):242-3.