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.)
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.]