From the Princeton Election Consortium
By Sam Wang
April 6, 2014
I’ve received some useful feedback on (my) NYT piece on autism. I have extra notes to offer on three topics: (1) elective induction, (2) environmental toxins, including endocrine disruptors, and (3) the lack of a true link with SSRI use.
Elective induction. In my last-minute editing, I joined two ideas that are really separate. The first is the risk of premature birth for autism. The second is the risk associated with early induction or unnecessary cesarean section. It is best to think of these categories separately.
The reasons for prematurity are not known, and are likely to include genetic contributions. A number of risk factors have also been identified, including tobacco and alcohol use. But in either case, premature birth is not the same as inducing delivery or cesarean section. Those are usually done after 37 weeks, a time that is not defined as premature.
My point was that sometimes these procedures are done even when there is no medical reason – the definition of “elective.” Such procedures carry a variety of risks to the baby, including autism. Fortunately, elective deliveries have fallen dramatically since 2010 (and here’s a state-by-state breakdown). Anyway, keep that bun in the oven if you can!
Endocrine disruptors and other pollutants. In the table we had an entry for “air pollutants, including mercury.” Endocrine disruptors fall into this category, and carry a risk ratio of about 1.3. However, these chemicals might not be the actual cause.
As I pointed out, pollution is higher in densely-populated areas and is associated with poverty, factors that are associated with stress hormone signaling. Stress hormones (and other aspects of the stress response) are known to affect brain development. So it could be the stress response itself that mediates any observed effects.
Consistent with this idea is the fact that so many pollutants have been linked statistically to autism, almost always with small, similar levels of increased risk. Such a nonspecific association makes me suspect that the direct biological cause isn’t the pollutant itself, but some correlated variable like stress.
Lack of risk from SSRI use. In the graphic I listed use of serotonin-selective reuptake inhibitors (SSRIs such as Prozac and Paxil) as having no statistically significant effect on autism risk. Several years ago came a report of increased risk. One difficulty with such studies is the fact that SSRIs are usually taken for a mood disorder such as depression.
Mental illness is heritable, and shares some of its genetic causation with autism. So if an SSRI user is more likely to have a baby with ASD than a non-user, it could be genetic rather than caused by SSRIs.
However, it’s hard to fully identify genetic risks, since both linkages between disorders (i.e. depression and autism) and specific genes have not all been discovered yet. As a result, indirect controls are necessary. One of the better controls is comparing SSRI use during pregnancy vs. SSRI use in the 2 years before pregnancy. In the second case, SSRIs don’t linger long enough to expose the fetus. This comparison shows no difference.
An even better control is comparing sibling pairs in which only one was gestating when the mother was on SSRIs. Here the risk ratio is not statistically significant and is 1.1, quite close to 1.0.
There may be other reasons to be careful about SSRI use during pregnancy. But for now, these reasons do not include autism risk.