It is not the drugs fault?

A frequent commentator, jgarveyrose, asked me if there was an association between psychotropics and homicide. This is a very hard question to answer, because the rates of homicide are low, and mass murder lower. The addition of information to such a low baseline rate, according to Bayes’ theorem, means that the increased risk is of no real predictive value: if the rate of homocide is 1 in 10000 and a factor reduces it to 1 in 1000 you would be intervening in a large number of people to save one event.

The one paper I can find is a decade old and dutch. The rates of suicide and homicide are noted in this table, as are the rates on the same scale of antidepressant prescriptions.

Graphically and over time, it looks like this.

 

There was an association between antidepressants and homicide rate. It was negative one: in the Dutch population, increased antidepressant use was associated with a decrease in homicide rate.

From the paper:

To our knowledge we are the first to report on the relation between lethal violence directed towards others and the use of antidepressants over a long period of time (15 years). We found a significant negative association between homicide and prescription of AD and SSRI ± Ve in the Netherlands, indicating that in a period in which the exposure of the Dutch population to antidepressants increased, the homicide rates decreased. As is the case for suicide, the association is strong, changes in the use of antidepressants can account for 65% of the variation in homicide rates. These data lend no support for an important role of antidepressant use in homicide. With regard to the subgroup analysis, the younger male groups are of primary interest. Here, the same negative associations are found.

These findings are in contrast with a recent study based on reported adverse events in the Adverse Event Reporting System of the FDA from 2004 to 2009, which concludes that “acts of violence towards others are a genuine and serious adverse drug event associated with a relatively small group of drugs [including] antidepressants with serotonergic effects”

There is less data on mood stabilizers and antipsychotics, but what there is seems to indicate that nonadherence to medications, not taking medications is the risk factor for homicide. This fits with clinical lore: we worry when people come “off their meds”.

As seen in Table 2, congruence between dispensed prescriptions and positive toxicology for antidepressants did not significantly affect the risk of homicide offending (aOR = 1.2; 95% CI, 0.5–3.1) or homicide victimization (aOR = 1.2; 95% CI, 0.7–2.2); by contrast, incongruence (here, at least 1 dispensed prescription in the absence of positive toxicology) was associated with a significantly increased risk of homicide offending (aOR = 6.2; 95% CI, 3.3–11.6) but not homicide victimization (aOR = 0.8; 95% CI, 0.3–2.0). Similarly, incongruence between dispensed prescriptions and positive toxicology for antipsychotics and mood stabilizers was associated with a significantly increased risk of homicide offending (aOR = 7.0; 95% CI, 2.8–17.7), whereas congruence did not significantly affect the risk of either homicide offending (aOR = 3.0; 95% CI, 0.9–10.5) or homicide victimization (aOR = 2.3; 95% CI, 0.9–5.8). In summary, nonadherence to medications used to treat affective and psychotic disorders elevated the risk of homicide offending.

There will be many anecdotes, but the plural of anecdote is merely case series. These large cohort studies are the best sets of data we have. So, in general, and despite lawyerly sophistic weaselling, there is limited to no evidence that provides a causal link between the use of psychotropics and violence. There may a link between stopping medications and violence.

4 thoughts on “It is not the drugs fault?

  1. The expression level is too low to ever be more than noise within a larger data set. Epidemiological surveys & studies can be quite useful, but they’ll never have refined data.

    Though you’d never catch the issue with a large sample size. Police treat non-drug related mass shooters as suicide missions by the perp, because that is what they are. At first response with force, they’ll almost all turn the gun on themselves.

    This goes back to my interesting break down of Suicidal Ideation some months ago. The anti-depressants shift the response criteria of the body & mind, suppressing some of the physiological effects of the depression. However, the shift seems to come in what the form of suicide ends up being. Rather than a normal, solo suicide, we’ve seen upshifts in Murder-Suicides, Suicide by Cop and the biggest one, Suicide after Mass Shooting.

    The problem is that since Suicide is a “sticky” though, bigger events puts the concept into the minds of the suicidal. As an American, the Columbine massacre wasn’t some crest of a wave of school shootings. It was the actual start of the wave. The wall-to-wall news coverage meant that the idea was stuck in the mind of every suicidal teenager in the country, which is why so many copy cat events happened.

    So it’s a split effect. A slight decrease in homicide rate with medications should make sense (people use rage for energy when they have nervous system issues, and it eventually overtakes them), but it’s shifted the profile for the extreme outliers.

  2. The question is “who is funding those studies?”. Studies like these cost a lot of money and those paying for them typically don’t want published anything that might reflect poorly on their products or agenda (climate change is a prime example). Would say that the current state of published papers is such that almost none of them can be trusted and would lump this into that category.

    • Check the papers. I did link to them. This is epidemiologists using large datasets and research grants.
      Most of the time there is no pharma money here: the pharmas have enough trouble funding RCTs for their new products.

  3. It’s a matter of framing the question in the proper form.

    The proper question to ask about causality in this kind of incident is: Would Mr A have attacked Mr B if he hadn’t been taking drug X?

    It’s not that the drug makes Mr A attack Mr B in such a time and place and using a specific method; but that if Mr A *hadn’t* taken the drug, then he *wouldn’t* have attacked Mr B.

    This kind of thing can often be established with legal (and reasonable) levels of certainty – indeed it is a routine mode of common-sensical thinking. Everybody knows that a drunk may do things the same person would not do when sober – and this can be (has often been) established in specific cases.

    For example, a great deal of violence (probably about half the stuff the police see) would not have occurred without alcohol.

    For some reason (actually, I think I know the exact reason! – see Pharmageddon by David Healy or Anatomy of an epidemic by Robert Whitaker) we do not apply the same common-sense reasoning to prescribed psychotropics as we do to unprescribed psychotropics (such as alcohol or social drugs).

    If an airline pilot causes an accident or deliberately crashes a plane when on alcohol or heroin, the drug is assumed to be significantly responsible – but if he does the same on a prescribed agent (which affects the mind at least as much as alcohol) then the standards of proof, the whole framing of the issue, is totally different.

    What we do have is significant epidemiological evidence that alcohol use is associated with adverse effects (and are some drugs — not all). The problem is that once something bad has happened we then look back and find causes. The Bayesian post test probability is 1.00. When I’m asked to predict, however, which drunk is going to crash… I get twitchy.

    And now that a *vast* number of people in developed nations are on prescribed psychotropics (often highly multiple – quite often five different drug classes – e.g. anticonvulsants, benzos, antipsychotics, psychostimulants and antidepressants) individually affecting many aspects of behaviour and also interacting in unknown and probably unknowable ways); we are seeing more and more problems.

    Agree. Am working on guidelines — watch this space!

    And previously rare problems (such as suicides and shootings caused by SSRIs, antipsychotics and other antihistamine-derived drugs with the class effect usually termed ‘akathisia’) have therefore become common.

    Akathisia was very common with typical antipsychotics, and I’m seeing a fair bit with the newest one (aripiprazole). It is less common with SSRIs but since so many people are on them… law of large numbers… we see a bit. The causal loop however, is not proven for the (fortunately rare) tragedies such as suicide and homicide. Bruce, do you have any papers on this you can email me? I did not do a systematic search — I save those for papers, not the blog

Leave a Reply