Recently a link was making the rounds on Tumblr about how SSRI anti-depressants caused violent and homicidal reactions in people (h/t to the lovely Cara for making sure we saw it). I was largely ignoring it because, frankly, there’s a lot of unproductive discussions about whether SSRIs, or anti-depressants in general, or even psychotropic drugs as a whole, are teh most awesome things ever! or an evil tool of big pharma or poisoning our children or should be put in the water supply to help the population at large. And my attitude towards psych treatment, whether it be therapy or medication or anything else, is pretty similar to my attitude about religion: everyone has the right to make their own determinations about their treatment and whether they would or would not like to take psych meds, and just as my atheism doesn’t make someone else’s faith any less valid, I can support someone’s decision to reject psych meds without lessening my own right to believe they help me personally. (To extend that, I support anyone’s right and decision to pursue or not pursue any kind of treatment, to identify as they feel appropriate, and to reject the whole framework and basis of psychiatry.)

That said, I thought it might be useful to take a look at this article to discuss some of the issues I see in a lot of these arguments and discussions. The basic gist of the article is to publicize an archive of “3,500 crime related news reports linked to the use of SSRI antidepressants … Pharma and the FDA may still be agnostic about SSRIs causing violence but 700 murders, 200 murder-suicides and 47 postpartum depression cases, including the 2006 case of Andrea Yates who drowned her five children on Effexor, don’t lie.” The article describes the site as “more of a public service than the FDA which has yet to withdraw the drugs named in the 3,500 stories–or even call them dangerous,” so the clear goal of the article is to encourage prohibition of SSRIs as a class.

That’s a pretty broad goal to be supported by such thin and unconvincing evidence – and that’s my problem with these kinds of arguments. Whenever I talk about these kinds of science articles, I often use the same phrase: “correlation does not equal causation.” This means that although two variables may be very closely associated, there’s not enough information to figure out which of them causes the other, or even if the two are related by anything more than chance and coincidence. A simple example is that everyone who orders food at McDonald’s is a human. Can we assume that if a dog walked into McDonald’s and ordered food, it would magically transform into a human because of the correlation between ordering at McDonald’s and being human? No.

To unpack this further, let’s look at the fact that a lot of people who walk in my neighborhood have dogs. I know some people who got dogs specifically in order to encourage themselves to do more walking – so the dog is influencing how often they walk. However, I also know people who do a lot of running or hiking and got a dog to keep them company on their outings – so their walking/running influenced their having a dog. So does being a person who walks outdoors make you more likely to get a dog, or does having a dog make you more likely to walk outdoors? We just do not have enough information to figure that out. This means my observations of walkers and dogs should not justify a public policy to issue dogs to every household to ensure people walk outdoors.

To extend this to the SSRI stories, there’s not enough evidence for us to determine if people who are not violent or homicidal become so when they are given SSRIs, or whether people who are already violent or homicidal are likely to be given SSRIs for treatment. And that’s a very important thing to be absolutely clear on when we’re talking about having the FDA eliminate an entire class of anti-depressants that some (including me) rely on for treatment.

There’s a couple other factors in the dog analogy that I also see at play in this SSRI story:

• Observer bias: I think dogs are pretty cute, so when I’m out and about, I tend to notice pedestrians with dogs more than I do pedestrians without dogs. So if I see 10 pedestrians and 4 of them have dogs, I’m much more likely to notice and remember the dog people and think that the majority of pedestrians have dogs. Similarly, self described “anti-SSRI advocates” are more likely to notice and prioritize instances where SSRIs occur with violent behavior.

• Who is observed: I happen to live two blocks from the most popular dog park in town, so people from miles away drive here in order to hike with their dogs. There are a lot of trails that don’t allow dogs and if I lived right next to one of them, I’d likely see a lot more walkers who are dogless. So I’m not looking at the entire population of walkers and dog owners when I’m observing a connection between those two characteristics – I’m looking at a population more likely to suggest to me that the two are connected. Similarly, the SSRI stories are drawn entirely from crime reporting. Stories about people who take SSRIs and do not engage in violent, homicidal, or otherwise criminal behavior are not going to be in a crime story – so the archive is looking at a subset of SSRI-takers that is more likely to confirm their perception that SSRIs cause criminal behavior.

• Interpreting evidence to fit desired results: when growing up, I tried out the “dogs will make me walk and exercise more” argument on my parents. This is because I wanted to get a dog, and I was trying to put together any argument I could to support that conclusion – I had started with the conclusion instead of with the evidence. Some of the stories mentioned in the article make me wonder if the SSRI stories suffer from the same problem. One of the quoted stories is “Lynyrd Skynyrd harmonicist Mike Caruso’s remark that, ‘the doctor put me on Cymbalta. That turned me manic.'” To me, giving an anti-depressant to someone with undiagnosed bipolar and triggering manic behavior is a very different argument than if taking the SSRI created violent or homicidal behavior that hadn’t previously existed in the person.

In order to convince me that SSRI used caused these behaviors in people who otherwise would not display them, I would want to see a clinical study where people were observed before and after starting SSRI treatment and a control group was also monitored while not having SSRI treatment. Those kinds of scientific studies are the only way to meaningfully determine whether the two variables have any kind of causal relationship.[1] And without that data, this article does a lot more harm than good – by reinforcing existing perceptions that criminals are all mentally ill and by shaming or scaring people who take and benefit from SSRIs.

[1] I should note that requests for data and scientific studies are often used to invalidate or minimize reported personal experiences from marginalized groups, an academic privilege argument of sorts. I do and continue to credit individual experiences where SSRI treatment caused specific behaviors for that individual, but I feel very uncomfortable making blanket decisions about whether or not these drugs should be available at all, for anyone, based on third party descriptions of the experiences of others.