In November 2013, Sorensen and colleagues from Denmark published a study looking at 668,468 births in Denmark from 1996 to 2006. The rate of exposure to antidepressants during pregnancy is very low in Denmark (about 1.3 percent in Denmark in this study versus 5 percent to 15 percent in the US) but what these researchers found regarding SSRI exposure was concerning. From Table 2 in their paper, one can see that the Adjusted Hazard Ratio for autism spectrum disorder in the children exposed to SSRIs was 1.6.
What this means is that they found a 60% increase in autism in the children of the moms who were using SSRIs during pregnancy. One way to tell if one thing causes another is to look for a dose –response relationship. For example, if smoking causes lung cancer, then we would expect to see heavier smokers have more lung cancer (and this is, in fact, what we see with smoking.) In this Sorensen study, the researchers looked at whether there was a dose-response relationship with the SSRIs during pregnancy and autism and they found that there was. Women on a low-dose of SSRI had a 40 percent increase in autism (aHR=1.4), while those on a high-dose had an 80% increase (aHR=1.8).
These results are very concerning and correspond to the prior human studies (as well as the theory and the animal studies presented above.) But, what the authors also did in this study was to take their initial population of 668,468 and then basically “throw out” more than 662,000 of the study subjects in order to do a substudy that focused only on 6,080 children of moms who had been given a diagnosis of an affective disorder through an in-patient or outpatient psychiatric hospital unit. When the authors did this much smaller study (about 100 times smaller) they still found SSRIs to be associated with autism spectrum disorder in the offspring but the result was no longer statistically significant. But this loss of statistical significance should come as no surprise given that they removed about 662,000 subjects from their original study.