A new study refutes prior research linking epidurals with autism spectrum disorder.

You would be hard-pressed to think of a medical innovation that has alleviated more human suffering than epidural analgesia.

Epidurals changed the process of childbirth from what, for many women, was an agonizing, if rewarding experience to one that was, well, manageable and rewarding. In the US, 73% percent of births are to women who have received an epidural. It is not only a common practice, it is the norm.

That epidurals might increase the risk of autism in a child, as reported in an October study in JAMA Pediatrics out of the Kaiser health system, casts a pall over the practice, and the headlines quickly followed.

The response to that original article was pretty swift, with the major pediatrics, anesthesia, and ob/gyn societies among others denouncing it as flawed, pointing out that there was a failure to adjust for multiple potential confounders.

As an aside, I want to point out that epidurals are about more than avoiding pain – not that that isn’t a noble quest – they also allow individual titration of analgesia and, should an emergency develop, the ability to do a painless c-section within moments, which may save the life of both mother and baby.

The problem with the Kaiser study was that, if true, pregnant women would face a choice between their own comfort and the potential health of their child. And of course, as any mother will tell you, the health of a child will come first.

But a new study, also appearing in JAMA Pediatrics came to a different conclusion, suggesting that the use of epidurals has no link to autism spectrum disorder in children.

What's going on and what should we believe?

In situations like this, it's important to first consider biologic plausibility. There are some key facts about autism that can help us judge whether it's even reasonable to think that epidurals could be a risk factor. For one, the monozygotic twin concordance rate of autism is up to 90%, whereas it's just about 10% in dyzygotic twins. This is fairly strong evidence for an underlying genetic disposition, which should obviously be independent of epidural use.

Environmental exposures may also be a factor, but neurodevelopmental studies have mostly found the first trimester, where complex brain development is accelerating, to be the most sensitive to environmental factors that may predispose to autism. And, obviously, epidurals don't take place in the first trimester.

But a link between epidural use and autism is not completely out of left field. Could small doses of the anesthetic or low dose opiates enter the mother's bloodstream and then the child's brain? Could epidurals permit longer labor times when, in their absence, a woman might have already proceeded to C-section?

 The new study helps us untangle these issues. Leveraging multiple large Canadian provincial databases that we in the US can only dream about, researchers were able to identify 123,000 singleton births. Just under 40% of them were born to mothers who received an epidural.

The raw numbers were similar to what was found in the Kaiser cohort. 2.1% Of the children born to mothers who received epidurals were later diagnosed with autism spectrum disorder, compared to 1.7% percent of those who did not receive an epidural- that’s about a 25% increased risk.

Of course, mothers who receive an epidural are different in multiple ways than those who do not.

I've outlined some of the key differences here. You can see that mothers who received epidurals tended to be younger and more highly educated and were more likely to be delivering in a tertiary care center. You can also see that they were more likely to have fetal distress – remember that for a second, I’ll come back to it.

The researchers did a series of analyses to account for these potential confounders. As more and more of these factors were included in the statistical model the association between epidural use and autism spectrum disorder decreased until it became non-significant.

The conclusion? The observed relationship between epidural analgesia and autism in children is not causal, but an illusion due to the fact that other factors associated with autism are also associated with getting epidural analgesia. Classic confounding.

I think that’s probably true, but I have to mention one issue since it is one of my ten commandments of statistics: don’t adjust for things on the causal pathway.

Remember that issue of fetal distress? One way to look at this – the way the authors did – is that fetal distress might prompt an epidural to be performed. And fetal distress may be an indicator that the baby will have problems (including autism) later in life. That makes fetal distress a confounder.

But we don’t really know that’s the way this worked. If an epidural led to fetal distress, which then led to some type of brain injury that manifested as autism, adjusting for fetal distress would be inappropriate – we don’t adjust for things that lie on the causal pathway. If we do, we ablate a true causal association.

Now – to be clear – I don’t think this is an issue here. Epidurals don’t seem to increase fetal distress in other studies. And the other perinatal factors – like fetal macrosomia, augmentation of labor, and the like – all feel more like confounders than causal mediators to me. But I needed to say it, because we need to think about this stuff and never just adjust willy-nilly.

Fortunately, the authors clear up the water I just murked by doing a sibling analysis. With knowledge of family structures they could compare the autism rate among siblings in the same family, where one was born with epidural analgesia and one was born without. This within family analysis accounts for factors that would be more difficult to measure from their databases, but would be stable within a family. This analysis also showed no difference in the risk of autism spectrum disorder among those siblings born after an epidural versus those born without.

So it seems that we can explain the findings from the Kaiser data as just another example of unmeasured confounding in observational research. The Canadian paper adjusted for more stuff, and there may still be more left out there to adjust for. One thing I keep thinking about, but is really hard to measure, centers around access and faith in the healthcare system.

It sounds a bit obvious, but in both studies kids were identified as having autism spectrum disorder if they had a diagnosis of autism spectrum disorder. But to get that diagnosis requires a few things. It requires access to a doctor who can make that diagnosis. More subtly, it may require some advocacy on the part of a parent who thinks something is not developing typically in their child. They might have to visit several specialists before they obtain that diagnosis. Could the type of mother who advocates for their own child that way also be the type of mother who advocates for an epidural for herself? If so, it seems like a reasonable explanation for some of the observational findings.

I don't envy pregnant women these days. The proliferation of large databases has made observational research more vigorous than ever before but has also added innumerable new potential risk factors to worry about. A pregnant woman who tried to do everything observational research told her to do would end up eating basically nothing, simultaneously taking and avoiding the same vitamins, and somehow having a fully natural water birth at home while in a dedicated labor and delivery unit at a sophisticated academic hospital. 

For now at least, let's not add epidurals as yet another source of mom guilt.

A version of this commentary first appeared on medscape.com.