Health and Science News for Parents

That link between autism and labor induction/augmentation? Yea, take another look

written by Tara Haelle

I don’t often blog on autism. It’s not my specialty, though it is an area of interest, and it’s something I’m still learning about. (I have a big post planned on autism misconceptions in general, but it’s still percolating in my brain.) However, I often write about the studies that come out linking autism to this, that or the other (or not linked to this, that or the other). And after such a study was published today, I simply could not stop myself from blogging about it because it’s just. so. dumb. The study suggests that kids are more likely to be autistic if their mothers’ labor was induced or augmented during birth. (Nevermind the fact that an induction or augmentation may have been medically necessary to save a woman’s life, address a complication like pre-eclampsia or avoid a C section.)

But before I dig into the study’s weak findings and myriad limitations, first consider that everything plus the kitchen sink has already been “linked” to autism (despite the strong genetic link for autism). Just a partial list includes air pollution, mom’s antibodies, mom’s depression, low birth weight, high birth weight, being born in the summertime, fertility drugs and living near a highway. But one more possibility won’t hurt to consider, right? (Except that it might divert research funds being used to find these links from more useful research, such as effective interventions for autistic children, but I digress.)

Autism has been linked to so many possible "environmental" influences that it's hard to keep track -- or to keep track of how weak the link might be. Photo courtesy of Microsoft images.

So, the study. It’s large, with 625,042 births pulled from a North Carolina birth database, including 5,500 kids with autism. Big numbers in a study are good. The mothers were between the ages of 15 and 49, and the kids had a birthweight of at least 400 grams. Wait — what? I assume that’s a typo (400 g is 0.88 pounds, or about 14 oz), but it’s unlikely that they meant 4000 g (8.8 pounds), so I don’t actually know what the kids’ minimum birthweight was for inclusion in the study. (UPDATE: I heard back from the authors, and the 400 g is correct. The minimum weight for inclusion was 400 g or 14 oz.) HOWEVER, the researchers evidently had access to data on the kids’ birthweight, so keep that in mind.

The study is retrospective, which means the data was gathered after the fact rather than tracking the pregnant mothers and their children forward. Many studies are done this way, but they are less reliable and informative than prospective studies that follow a population forward. With retrospective studies, the only data available is the data that was collected at the time. If there are other possible confounders (factors that could affect a person’s risk of something and interfere with your assessment of another risk for the condition), there is no way for the researchers to gather this information or to interview the mothers to see what other differences might have existed among them.

Another problem with this particular study’s being retrospective is that identification of autism in the children was done only by looking at their school records. If they had a special education “exception” for autism (diagnosed by a clinician and further evaluated by a school psychologist), they were included as autistic. The problem, noted in the study’s limitations, is that if the kids had a special education exception for something else, such as a learning disability with only a secondary diagnosis of autism, they were not included in the autistic group. There is no telling how many kids this might apply to, but it’s likely that at least some of the kids in the “non-autistic” group actually have autism and were missed because it wasn’t the primary reason for special education on their school records.

Let’s assume for now that there aren’t enough of those missed kids to make a difference, though, and look at all the factors the researchers took into account in calculating whether a child’s risk of autism was higher if their mother had an induction or labor augmentation.

They controlled for the child’s sex, the child’s birth year, the mother’s age and race/ethnicity (black or white only; Hispanics were excluded), whether a child was first born, whether the child was a single or multiple, the mother’s education and the mother’s marital status.

They did not control for father’s age, one of many factors linked to a higher risk for autism, because the info wasn’t available. (Again, those blasted retrospective studies.)

They also controlled for type of birth (vaginal, C section, unknown), any type of hypertension or diabetes in the mother (including pregnancy-induced forms), gestational age (prematurity) and smoking during pregnancy (which may or may not be a risk factor for autism).

Finally, they controlled for various complications during labor and delivery: newborn fever (which could indicate bacterial infection), meconium aspiration, breech presentation (feet first instead of head first), fetal distress, placental abruption and cord prolapse (umbilical cord precedes the baby during birth). Meconium aspiration and breech presentation are supposedly linked to autism in past studies, but the link to meconium was in a very small study (only 91 children) and the reference to breech presentation was a meta-analysis that does not even mention breech in the abstract.

So, they considered a lot of stuff, huh? Where is birth weight?

They adjusted for type of birth, prematurity and all sorts of birth complications — yet not birth weight? Or head circumference? These would seem to be incredibly obvious things to control for given that a large baby or large head circumference are both sometimes used as medically indicated reasons to induce or augment a woman’s labor — and that large head size is linked to autism. In fact, about 20 percent of people with autism have a head circumference above the 90th percentile. Since a large head is a pretty strong overlapping factor between autism and mothers who are induced or have labor augmentation, why would the researchers not take it into account?

UPDATE: I emailed the authors to ask them why neither birthweight nor head circumference were included as confounders. Here is the reply from corresponding author Simon Gregory, PhD, who said he checked with his colleagues and provided these comments:  “At least clinically, of the two, birthweight and gestational age, gestational age is more predictive of how kids end up doing, at least in the neonatal period. The minimum birth weight for inclusion in the study is at least 400 g. Head circumference is not measured in the detailed birth record. While we did not control for birth weight, we controlled for gestational age, which is highly correlated with birth weight.”

So, unsatisfying response, but moving on, let’s look at what the authors actually found.

In terms of raw numbers (before any adjustments for other factors are made), 31.8% of the autistic boys’ mothers had been induced and/or augmented, compared to 28.7% of the non-autistic boys who had been induced or augmented. The numbers for girls are even less impressive: 28.3% of non-autistic girls and 29.4% of autistic girls were birthed by mothers whose labor was induced and/or augmented.

Using this information and controlling for nothing other than the children’s sex (autism is diagnosed four times more often among boys than among girls), a child was 1.23 times more likely to have autism if his/her mother’s labor was induced or augmented. (That’s not even twice as likely.)

After controlling for everything else, the risk for autism was 1.2 times* greater with induction and augmentation, 1.1 times greater for induction only and 1.14 times for augmentation only. Note that 1.2 is not much lower than the unadjusted 1.23 times greater risk, so all those adjustments they made don’t appear to have made much difference in the children’s risk — despite the fact that their analysis found a higher risk for autism if there was meconium aspiration (1.22 times), fetal distress (1.25 times), preterm birth (1.25 times for 34 weeks or earlier) or maternal diabetes (1.23 times).

All of these very tiny increases in risk for autism were statistically significant, which means they do not appear to be due to chance. But consider this other association (also statistically significant) included in the study’s data tables: children of mothers with some college – after controlling for all the same factors – had a risk 1.09 times greater for autism than the children of mothers with a high school education. For children of mothers with a college degree, the increased risk was 1.31 times greater for autism than for children of moms with a high school degree.

So, if you’re a kid whose mom has a college degree, your odds of being diagnosed with autism are GREATER than your odds of being diagnosed with autism if your mother’s labor was induced AND augmented. Now, there are many reasons a mother with a college degree might be more likely to have a child diagnosed with autism: she may be more educated about the symptoms or have more resources available for identification/diagnosis. But the fact of the association reveals how unreasonable it is to attribute any associations in this study — including labor interventions — to a “cause” of autism.

Meanwhile, consider what the authors explain they did not control for because they didn’t have the data (though again, oddly, they don’t mention birth weight or head circumference): any medications the mothers took during pregnancy or during birth (not that any are linked to autism, right?), any labor abnormalities (including ones that, oh, might indicate the need for augmentation), the mother’s pre-pregnancy weight, father’s age (already linked to autism) and siblings (a sibling’s autism diagnosis increases one’s risk of autism).

Yet, despite all these limitations and the small risk increases, the researchers calculated that “if male exposed children theoretically became unexposed [no labor induction/augmentation], 2 of 1,000 children would no longer have a positive autism diagnosis.” Whoa! If we eliminated all use of induction and augmentation (nevermind the medical emergencies that can require either), then 2 fewer boys of every 1,000 would have autism? That implies that the induction/augmentation caused the autism, which this kind of study cannot show at all.

The authors try. They suggest that exposure to “exogenous oxytocin” (AKA pitocin) might be one explanation for the link. The problem? The study cites another paper that discusses whether it is theoretically possible for oxytocin to influence behavioral outcomes based on sheep, voles, rats and mice. Then these study authors state, “Exposure to exogenous oxytocin during induction/augmentation may have a functional effect through, as yet, unidentified genetic or epigenetic factors.” In other words, pitocin might somehow cause autism, but we have no idea how.

The authors do note the “significant maternal and fetal benefits of labor induction and augmentation including reduced fetal/neonatal death and meconium aspiration syndrome; lower cesarean delivery rates; lower risk for neonatal ventilation, sepsis, and intensive care nursery admission; and reduced maternal mortality.” But the possible benefits of induction/augmentation for some women do not stop the authors from coming to the fear-mongering conclusion that induction/augmentation might lead to autism. Given the publicity this study has received, how will this affect a woman whose doctor tells her she medically requires an induction or augmentation?

To recap: we have a retrospective study showing a tiny increase in risk for autism among kids whose mothers’ labor was induced or augmented. The risk does NOT take into effect the father’s age, mom’s medications during pregnancy, the size of the baby’s head or the baby’s birthweight. The increased risk with induction/augmentation is LESS than the risk for autism if the mother has a college degree. And they offer no method by which labor induction or augmentation (or the medications used for them) might actually cause autism (though that doesn’t stop them from saying that not inducing/augmenting could somehow theoretically reduce autism cases by 2 out of 1,000).

Yep, I’m convinced.

UPDATE: Check out Emily Willingham’s analysis of the study, in which she points out several flaws I missed.

[All these numbers are odds ratios, not the greatest ways to explain/understand risk, but often the best statistical info provided. For reference, an odds ratio of 1 means there is no greater risk, and an odds ratio of 2 means the risk is twice as great in the exposed group than in the comparison group.]

28 Responses to “That link between autism and labor induction/augmentation? Yea, take another look”

  1. Darwy

    That’s one of the things I noticed when I read the study today – they didn’t break down the prevalence of autism by birth weight – something which, in my opinion, is a MAJOR flaw in this paper.

  2. Alan Duval

    Ignoring for the moment that they’ve missed the correlation/causation issues that may be involved (the missing head circumference data and a mental development issue that has suggested physiological markers), is there any data on phenotypic traits in the parents?

    • Tara Haelle

      There is none included in the paper. Because it’s retrospective, it’s likely they did not have that information (yet another flaw).

  3. Did they control for type of induction? E.g., use of prostaglandins, Pitocin, articial rupture of membranes, etc? It seems like that would be relevant as well.

    • Tara Haelle

      I agree that would be relevant as well. No, they did not control for method of induction or augmentation. They mentioned the use of oxytocin (pitocin) as one possible mechanism for the association, presumably because oxytocin is likely the most common form of induction and/or augmentation. But they did not mention use of a cervical softener, artificial rupturing or any other methods. Their categories were “induced or augmented; induced and augmented; induced only; augmented only; not induced or augmented.”

      They did distinguish between vaginal birth, operational vaginal birth (I presume that means vacuum and/or forceps use) and C section, but adjustments for these did not appear to affect the odds ratios (or not by enough to show up in 3 significant numbers).

  4. This is an excellent response to a study which is being used totally unethically, in my opinion, to scare women away from birth interventions. Thank you for laying it out in an understandable way. I can’t believe they didn’t adjust for head size or type of induction. Oh wait – I can. Because most of the studies which link future problems to a mother’s delivery method or postnatal behavior have a tendency to ignore the obvious.

  5. April Humphrey

    I think the article (that I read) was simply stating that of the children with Autism, a high percentage of them were induced. For that reason, it has to be retrospective. I think they were only highlighting a correlation, not causation. If the attempt was to link it more directly in a non-retrospective way, the study would have to be very different of course. But it was a simple, matter of fact article stating that kids with Autism have a high percentage of being induced. Anything else is really just “reading into” the article more than what it meant, which really is just as detrimental as people assuming induction causes Autism ;)

    • Helen E Davis

      But almost as high a percentage of the normal (without autism) children were induced. All we can come up with is that a high percentage of births were induced…

  6. Michael Southeren

    Another confounding factor that also often appears in retroactive studies has to with the statistical analysis process used and ‘researcher degrees if freedom’. Quickly put, the more different ways you look at some data, the more likely you will find something ‘significant’ by random chance. There is a fix you gave to include to make the numbers actually relevant to the real world. Want to have a guess as to how much error NOT doing this makes in a result?

  7. Colin O'Shea

    The information presented in the study cannot differentiate between, Autism babies being more likely to require induction and babies getting induced being more likely to be Autistic. If it is true that Autistic kids are more likely to have larger heads, then one would presume the mothers are more likely to get CPD and consequently require augmentation. To not control for this variable seems to be a major flaw. You cannot presume to correlate gestational age with weight if the subset you are analysing has a significantly larger head size.

  8. [...] analyses. [ETA: They had a minimum cutoff for birthweight of 400 g, as science writer Tara Haelle reports here. She also points out how limited the application of these findings would be, even if they did show [...]

  9. does it matter the procedure that was done for induction?

    • Tara Haelle

      They did not distinguish between different methods/procedures. Induction could mean rupturing of the membranes or could mean certain pharmaceutical use, such as oxytocin (pitocin), among other possibilities. But they did not look at specific procedures, so it’s not clear what the “association” (weak as it is) would be specifically tied to.

  10. [...] analyses. [ETA: They had a minimum cutoff for birthweight of 400 g, as science writer Tara Haelle reports here. She also points out how limited the application of these findings would be, even if they did show [...]

  11. The statistics on mother’s with a college degree made me want to toss my computer through the window. That, in of itself, might have made an interesting article, but only if it was titled “College educated moms may be better at providing resources to diagnose and treat their autistic babies.”

    Not to go all tin-hat on you, but it really seems like the authors had an agenda.

  12. It’s like ice cream sales and sexual assault rates. Both go up around the same time, but both are associated with something else: the weather.

    Retrospective or not, it’s valid research, but it’s certainly not definitive or causal. Having worked on studies like this (for completely different health issues), I don’t doubt the author’s agenda. I think they’re fairly neutral. They’ve tapped a LOT of data to produce this analysis and used virtually all of it to produce what they did. This, to me, looks like the work of intense research geeks that just report what they observe without necessarily considering how nutcases may (mis-)interpret it.

    • Tara Haelle

      I love the analogy of ice cream sales and sexual assault. I haven’t heard that one before. Normally, I’d agree that it’s just researchers crunching numbers — if it were anything other than autism. But I’ve seen so many of these studies on autism that I’m less willing to disregard the possibility of an agenda on this one. That’s especially so because of their statement that autism in 2 boys of every 1,000 could theoretically be prevented if there were no induction/augmentation. An agenda-less researcher does not include that kind of statement in a study which clearly does not show any causation. The authors here do not address the possibility of an underlying link with head size, body size or other birth complications. To ignore that completely (especially when they had birthweight available) and to only suggest oxytocin as a possible mechanism causing autism is intellectually dishonest. Their limitations are written disingenuously. Sorry, on this study, I don’t buy it.

  13. We had three boys; two of the three were non-verbal autistic, very hardcore autistic. The middle one talked for all three, and THEN some. *Grin*
    #2–induced. (The speaking one)
    #3–induced. (The second autistic one.)
    None of them were induced early, the latter two at 39 weeks. Obviously genetics (although neither family had any history of such) played a much greater role than inducing labor. Like cancer, autism seems to be caused by a great many outside factors coupled with a genetic predisposition. I also think the levels of actual austism have shown a real uptick in time, that not all the new autistic children are just a change in methodology of reporting. But obviously I don’t think any one thing does it, and I’m sure vaccination did not cause it, having studied the literature. (In fact, that’s about the only thing I’m sure is NOT causing it.)

    • Tara Haelle

      I honestly don’t know enough about the changes in reporting and the past diagnoses of other conditions to know whether there is a real uptick or not, though I suspect — based on what I’ve read from trusted folks who do know their stuff — that it’s mostly changes in how we identify and classify. Regardless, I do agree the etiology is likely very complex, with heavy genetic influence but possibly not only that. While I know more research into the etiology is important, I do wish less research money would be spent on that instead of spending it on research for effective interventions.

  14. Robert Benner

    Well done, Ms Haelle! I am a Board Certified OBGYN and have to say what an impressive analysis you have provided here. I would add that the data in the birth registry records usually comes from birth certificate datasheets and is well known to be of terrible quality. Even items such as whether pitocin was given or not can be very suspect–the difference of whether or not a box on a form is checked by a doctor who has finished the delivery and just wants to get the paperwork done and go home. Thanks for a very insightful analysis.

    • Tara Haelle

      Thank you! I was not aware of that re: birth registry records, so there’s one more flaw to add to the list. I realize that studies like this are difficult to do, especially in getting all the data needed, but there’s a point past which the number of holes should preclude moving forward, I think.

  15. Lauren

    My sister has an autistic son. I have two sons who are not remotely autistic. Our uncle is autistic. Other than different dads, neither with family histories of autism, we had similar pregnancies and other variables. Except I had no birth interventions (why yes, it did tremendously suck.) My sister did the whole assembly line pitocin and spinal assembly line birthing. Our subsequent histories are similar enough, including the standard vaccinations. Her son was exhibiting signs by age 2, seemingly not correlated to vaccination. Food freakiness earlier and disconnection. I do not know my nephews genetics results, but I do know that my father, me, and sons show polymorphism in the oxytocin gene that often are correlated with autism. It is reasonable to investigate birth inducing hormones as possible triggers for those who are genetically vulnerable. I was told at the hospital that virtually all women got pitocin, because docs didn’t want to stay past 5 or ruin holiday weekends. Both of my children were born on holiday weekends at 2 and 3 am respectively. I was the only woman using a birthing room those weekends, and it was a major metropolitan hospital. Both my labors were long and the contractions were ungodly strong. Yet they kept wanting to push pitocin on me. Even if it is a small percentage who are vulnerable because of genetic reactions to exogenous oxytocin, it is still important to look at.

  16. [...] induction. I will not get into the weeds right now on that particular study (although here is one very critical take), but it got me thinking again about labor induction, a topic that I cover at [...]

Leave a Comment - Here's your chance to speak.(eMail will not be published)


Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 111 other subscribers

Shot@Life Join the Movement

Great Links

Tag Cloud