If you’re parenting in the modern age, one thing you should know for sure is that suffering is a signal that you’re a good parent. This was reinforced for me best in an essay I read several years ago in which someone noted that being an attentive parent sometimes requires you to hold your pee in for several hours. Because it might not be best for your kids if you took a break to use the bathroom. (Frankly I do not get this at all because in my experience children have no problem being in the bathroom with you.)
This starts during gestation, when being a good parent means giving up things you enjoy, like coffee, sushi and sleeping on your back. With this background, it was no surprise to me to read this new study on how epidurals raise the risk of autism. Being a good mother should definitely mean going through an un-medicated labor. Honestly, it’s really not that bad. As my husband can tell you, I only clawed a small portion of my face off to distract myself. It’s a small price to pay!
Still, I thought it might be useful to review this study for people just in case it turns out to, maybe, not be 150% convincing.
[Snark over; data review commence.]
What does the study on epidurals and autism say?
This study, which was published in JAMA Pediatrics includes almost 150,000 children born in California between 2008 and 2015. At the most basic level, the paper correlates use of epidural during labor with a later autism diagnosis. The ability to do this actually reflects a nice aspect of the data, which is that it comes from one integrated health system (in California). This allows the authors to link medical record data on epidural usage to later autism diagnosis for a large sample and without relying on self-reports.
The top line conclusion of the study is a 37% increase in risk of autism diagnosis with epidural usage. Put in perhaps more useful terms: the share of children in the non-epidural group with autism is 1.3%, versus 1.9% in the epidural group.
This study is not a randomized trial. The approach the authors take is to compare children of women who have an epidural to those who do not. The problem, of course, is that other things differ. You can check out all the differences in their Table 1, but I’ve pulled of the top of it below. You can see there are very notable difference in racial and ethnic make-up of the two groups. If you look at the whole table, there are sizable income differences, among other things.
Like with all studies of this type, it is extremely difficult — read, virtually impossible — to conclude that it is the epidural that is driving the differences as opposed to some other feature of the data. The authors do control for differences across women but, as always, what about the differences we do not see?
Indeed, for this reason this study has been pilloried. Several national associations, including ACOG, came out strongly to argue that the study was not causal and should not be used to influence practice patterns. In addition to these issues of differences across women, these organizations point out that there is no biological reason to think there would be a link here.
It will perhaps not surprise you that I agree. Observational studies like this are really difficult to rely on for causal conclusions. But I thought it might be interesting to dig a little deeper into what, actually, might be driving the results.
Option 1: The link is due to other conditions
In Table 1 of the paper, in addition to these differences across baseline characteristics, we also see differences in pregnancy and infant complications. For example, children in the epidural group are 1.3 percentage points more likely to have a birth defect (they do not say what kind). Moms in the epidural group are more likely to have pre-eclampsia and diabetes before pregnancy.
What this tells me is that — on average — this is a group with more complicated pregnancies. Without more details about these pregnancies, it will be very hard to fully control for these differences.
Option 2: The link is due to differences in autism diagnosis
A basic fact from their Table 1 is that women with high socioeconomic status are more likely to have an epidural. For example, a much larger share of the epidural group has a yearly household income over $90,000. My hunch is that this means that differences in diagnosis rate drive a lot of these results. Families with more resources may be more likely to get an autism diagnosis, simply because they have more resources to access medical care.
Of course, it is also possible that there is a true causal effect here. I think if there is, it’s likely to be extremely small. The measured effect is already small, and after more careful adjustment, it would surely be smaller. Is it possible that there is some tiny causal effect buried in here? It is possible. But it it something to base decisions on? I would say no.
P.S. I am well known for being on the bandwagon of “observational studies are a problem”. However: I often feel I’m yelling into the void about this, especially in pregnancy. Studies of, say, caffeine are subject to basically the same critiques as the above. ACOG and others virtually never come out to say, hey, a lot of these restrictions are based on observational data which might be flawed. So I was heartened, I guess, to see such a strong negative reaction to this study. Still, it is perhaps too much to hope that we’ll see a more general push for causality.