I can understand the eagerness with which science writers jump on stories that deal with new findings about autism spectrum disorders (ASDs). After all, the mystery surrounding the rapid increase in ASD rates over the past 20 years has made any ASD-related study that may offer some clues inherently interesting. Because people are anxiously awaiting some explanation of this medical enigma, it seems like science writers almost have an obligation to discuss new findings concerning the causes of ASD.
The problem with many epidemiological studies involving ASD, however, is that we are still grasping at straws. There seem to be some environmental influences on ASD, but the nature of those influences is, at this point, very unclear. This lack of clarity means that the study of nearly any environmental risk factor starts out having potential legitimacy. And I don't mean that as a criticism of these studies--it's just where we're at in our understanding of the rise in ASD rates. After we account for mundane factors like increases in diagnosis due simply to greater awareness of the disorder, there's a lot left to figure out.
So, with all this in mind, it's understandable (at least in theory) to me why a study published last week in the journal Pediatrics became international news. The study looked at a sample of children that included healthy individuals along with those who had been diagnosed with ASD or another disorder involving delayed development. They asked the mothers of these children about their use of selective serotonin reuptake inhibitors (SSRIs) during pregnancy. 1 in 10 Americans is currently taking an antidepressant, and SSRIs are the most-frequently prescribed type of antidepressant. Thus, SSRIs are administered daily by a significant portion of the population.
Before I tell you what the results of the study were, let me tell you why we should be somewhat cautious in interpreting them. This study is what is known as a case-control study. In a case-control study, investigators identify a group of individuals with a disorder (the cases) and a group of individuals without the disorder (the controls). Then, the researchers employ some method (e.g. interviews, examination of medical records) to find out if the cases and controls were exposed to some potential risk factor in the past. They compare rates of exposure between the two groups and, if more cases than controls had exposure to the risk factor, it allows the researchers to make an argument for this factor as something that may increase the risk of developing the disease/disorder.
If you take any introductory epidemiology (i.e. the study of disease) course, however, you will learn that a case-control study is fraught with limitations. For, even if you find that a particular exposure is frequently associated with a particular disease, you still have no way of knowing if the exposure is causing the disease or if some other factor is really the culprit. For example, in a study done at the University of Pennsylvania in the late 1990s, researchers found that children who slept with nightlights on had a greater risk of nearsightedness when they got older. This case-control study garnered a lot of public attention as parents began to worry that they might be ruining their kids' eyesight by allowing them to use a nightlight. Subsequent studies, however, found that children inherit alleles for nearsightedness from their parents. Nearsighted parents were coincidentally more likely to use nightlights in their children's rooms (probably because it made it easier for the nearsighted parents to see).
A variable that isn't part of the researcher's hypothesis, but still influences a study's results is known as a confounding variable. In the case of the nearsightedness study, the confounding variable was genetics. Case-control studies are done after the fact, and thus experimenters have little control over other influences that may have affected the development of disease. Thus, there are often many confounding influences on relationships detected in case-control studies.
So, a case-control study can't be used to confirm a cause-and-effect connection between an exposure and a disorder or disease. What it can do is provide leads that scientists can then follow up on using a more rigorous experimental design (like a cohort study or randomized trial). Indeed, the scientific literature is replete with case-control studies that ended up being false leads. Sometimes, however, case-control results have been replicated with better designs, leading to important discoveries. This is exactly what happened with early reports examining smoking and lung cancer.
Back to the recent study conducted by Harrington et al. The authors found that SSRI use during the first trimester was more common in mothers whose children went on to develop ASD than in mothers who had children who developed normally. The result was only barely statistically significant. This fact combined with the variability seen in the confidence interval suggests it is not an overly-convincing finding--but it was a finding nonetheless. In addition to an increased risk of ASD, the authors also point out that SSRI exposure during the second and third trimesters was higher among mothers of boys with other developmental delays. Again, however, the effect was just barely statistically significant and even less convincing than the result concerning ASD.
So, the study ended up with some significant results that aren't all that impressive. Regardless, because this was a case-control design, there is little we can conclude from the study. To realize why, think about what other factors women who take SSRIs might have in common. Perhaps one of those influences, and not the SSRI use itself, is what led to an increased risk of ASD. For example, it seems plausible that the factors that make a mother more susceptible to a psychiatric disorder might also play a role in making her child more susceptible to a neurodevelopmental disorder. In fact, a cohort study published last year with a much larger sample size found that, when the influence of the condition women were taking SSRIs for was controlled for, there was no significant association between SSRI use during pregnancy and ASD.
The fact that this case-control study doesn't solve the mystery of ASD isn't a knock on the study itself. If anything, it's a knock on some of the science writing done in response to the study. I can't go so far as to say these types of studies shouldn't be reported on. But, they should be reported on responsibly, and by writers who fully understand and acknowledge their shortcomings. For, it is somewhat misleading to the general public (who likely isn't aware of the limitations of a case-control study) when headlines like this appear: "Study: Moms on antidepressants risk having autistic baby boys."
The safety of SSRI use during pregnancy is still very unclear. But both SSRIs and untreated depression during pregnancy have been linked to negative health outcomes for a child. Thus, using SSRIs during pregnancy is something a woman should discuss at length with her doctor to determine if treatment of the underlying condition poses more of a risk than leaving the condition untreated. In making that decision, however, the barely significant findings from a case-control study should not really be taken into consideration.
Rebecca A. Harrington, Li-Ching Lee, Rosa M. Crum, Andrew W. Zimmerman, Irva Hertz-Picciotto (2014). Prenatal SSRI Use and Offspring With Autism Spectrum Disorder or Developmental Delay PEDIATRICS DOI: 10.1542/peds.2013-3406d