Is an obese mom as bad as a drunk one?

Posted on June 7, 2017


Our starting point:

Association Between Prenatal Alcohol Exposure and Craniofacial Shape of Children at 12 Months

These researchers recruited pregnant women in the first trimester. At 12 months they used 3-d imaging to see differences amongst their kid’s facial shape based on alcohol consumption during pregnancy.

One symptom of fetal alcohol syndrome is changes in face appearance:

fetal alcohol syndrome changes in facefetal alcohol syndrome facial features

In this study we’re looking at children not diagnosed with fetal alcohol syndrome, but with moms who did consume alcohol. Perhaps 3-d imaging software can detect changes reminiscent of fetal alcohol syndrome (FAS) that a doctor wouldn’t catch?

  1. FAS is a spectrum, similar to autism. Where it’s not you have it or you don’t. You could have a different amount of it. Could there be some low level signature we haven’t been able to discern?
    1. This could be valuable in e.g. kids where there are developmental issues, but no apparent visual deformities. And perhaps some moms of these kids say they didn’t drink, but the imaging finds “Oh, this kid does have symptoms of alcohol exposure; mom may be lying to us.”
  2. Culturally we know heavy drinking is not a smart move while pregnant, but many are still under the impression a drink of wine a day is fine. Or a drink here and there is fine. The evidence on whether low level drinking is ok is murky. This study takes the subjectivity out, and gives us increased precision.

The levels of alcohol exposure were:

  • None
  • No more than two drinks in one sitting and no more than seven per week (low)
  • At least two but no more than five drinks in one sitting and no more than seven per week (moderate)
  • More than five in one occasion (binge)

It’s important to note a standard drink here is Australian sizes. We Americans out-do them in this department. (But Japan is king. Check out Wikipedia’s Standard Drink.) No more than two Australian drinks is no more than 1.5 American drinks.

The researchers broke this into tiers. Let’s look at a couple. First up, any alcohol exposure. This doesn’t tell us anything about the amount of consumption, just whether it happened:

On the left is the average difference by location. The further we get from dark blue, the greater the difference.

On the right, the difference is subcategorized by depth. Green is no difference, red means the face is sticking out more, blue means it’s receding. We can see the head is either normal, or the front of the face is indented.

This is noteworthy because one theory of fetal alcohol syndrome is the face does not grow outward like it should. There is potential to use this as a proxy for skull development, which means it could be a proxy for brain development. For instance, if the skull isn’t growing out as much as it should, it might not be as big as it should, where then the brain isn’t as big as it should be = > kid isn’t as smart as they should be.

Let’s get more precise. Perhaps the most relevant pregnancy drinking is having some in the first trimester, potentially because you don’t realize you’re pregnant, then ceasing all alcohol:

We can see this is looking worse. The differences are getting more pronounced, and all around the areas we’re concerned with for fetal alcohol syndrome.

What happens if the drinking was higher in the first trimester?

Even worse. Makes sense.

-> If you want to see all the different ways they looked at this, here is a full image.


1) We still don’t have a clear threshold. Remember our low intake group was no more than two drinks in a setting and no more than seven per week. One drink per week was thus in the same category as seven. That’s a big range.

2) If you look back at the numbers in the images above, you can see we’re dealing with millimeters. In many cases we’re talking about maybe a half millimeter difference. Is that relevant? Does a half millimeter become a full millimeter by adulthood? A half millimeter of less skull can be quite a bit less brain volume! What are the clinical ramifications of such a difference? Can this difference diminish through development through some type of treatment?

We don’t know!

That wasn’t the point of this study. The point was to assess are there differences? On average, yes, there are. But there was no neurobehavioral or any cognitive follow up. (They plan to do this down the line.)

3) Smaller skull size doesn’t mean less intelligence. Women have smaller brains yet are as smart as men. The brain is a lot more complicated than that. Plus, it can change over time.

3) There are a lot of factors which dictate appearance.

Even moms who drink their asses off don’t guarantee their kids fetal alcohol syndrome. According to a Wikipedia reference, only a third of alcoholic mother’s will have a kid with FAS.

The authors did their best to account for this. They used a statistical model aimed at handling tons of factors.

The partial R^2 is an attempt to garner the percentage each variable can explain the differences in craniofacial shape. You can see these percentages are quite small. Again, in some sense this is expected. If 70% of alcoholic mother’s don’t end up with deformed kids, that says a lot about how many and how important other factors are.

4) The results weren’t statistically significant, yet that’s ok.

I know this is eye rolling reading for many, but it’s crucial to go over.

This is our yearly reminder of what a p-value is. The p-values in this study were all over the place. A couple write ups made the (dangerous) inference that means the results were not reliable, so the data is ignorable. That is not how to interpret a p-value!

With health studies we typically set a p-value of .05. We want there to be at most, a 5% chance the results were due to chance. That we’re 95% or more sure of why we got our results. For instance, we’re 95% sure our sample was big enough.

This is an arbitrary value we set. It’s not some law of the universe.

If we get a p-value of .30, many results this study had were around this value, that means we can speak with 70% confidence. “We feel 70% sure our results were not due to e.g. random variations.” It doesn’t mean the results are suddenly garbage because they didn’t hit the p-value threshold. It just means they aren’t AS sure as we shoot for.

Say our hypothesis is there will be no difference between women who don’t drink and women who do. Then we get a p-value of .30. That means we’re not at our threshold for being able to say there IS a difference between women who drink and women who don’t. We can’t reject the hypothesis, but we can’t accept it either. Because that’s not how we set the experiment. A large p-value increases the *potential* reasons we got our results, but it doesn’t dismiss any reasons!

Our threshold for certain things are very high. Physicists are known to go out many decimal places. If you’re talking about a new drug, with nasty side effects, yeah, you want to be damn sure it actually hits the disease too.

We’re talking whether someone drinks or not during pregnancy. How confident do we need to be to tell wannabe moms you shouldn’t drink? When we’re pondering irreversible consequences, for an activity that looks to have zero benefit? Not very.

This is different than everyday adults. There is enough evidence suggesting moderate drinking may not only be fine, but for some is an activity they should keep doing. For instance, if it keeps you social, that can be a healthy activity. We’re not seeing that with pregnant women.

Even if the study said there’s a 90% chance the results were due to a screw up in the data, is that enough for anybody to say “Drinking is ok”? Of course not. Clinical interpretation isn’t binary. The risk of lung cancer in smokers in 10%, yet we’ve recommended nobody smoke.

One of the most important, underutilized phrases in research is statistical significance is not clinical significance. A study showing any hint of brain damage for an activity seemingly having no health benefits, is clinically relevant no matter the statistical significance.

The obesity connection

A logical assumption is “How many of these moms were also smokers? Or gained too much weight during pregnancy? What were the moms’ tolerance to alcohol? Age? etc?” The study controlled for many of these variables.

Though it was practically thrown away as extra data, they did an interesting twist. What if you reverse the variables? If you control for alcohol intake, what is the affect of BMI on craniofacial shape?

Not only can we see there is an influence, but it’s strikingly similar to the influence of alcohol intake, with indentation around the middle of the face!

And yep, the researchers controlled for birth weight. A heavier kid may have thicker cheeks. Indeed, that was found:

But if you control for that, you get a face (skull? (brain??)) which is no longer sticking out. Instead, it’s receding.

The effects of obesity on pregnancy are becoming a bigger and bigger deal. Risks of diabetes, preeclampsia, c-section, have all been implicated. Brain development is not something I’ve seen talked about much, if at all.

Sometimes research is only meant to provide a jumping point. This part of the study kind of got cast away. (Ironically, it had a statistically significant p-value!) There is some research already out there in animals-

Enduring consequences of maternal obesity for brain inflammation and behavior of offspring

Other areas have gotten less publicity too. Neural tube and cardiac defects being some.

As a new dad, I’ve heard a lot of talk about pregnant women and drinking. I’ve heard little about how heavy the average pregnant woman is. Should there be a similar level of fear with being obese as drinking?


I have a good deal of clients who have fully grown kids. My baby momma has a bunch of nieces and nephews all born within the last decade. As we went through our pregnancy process, we realized quite a bit of what was best practice for those parents was different for us.

In some cases, older parents do not like to hear this. It’s tossed away as new parents being unnecessarily cautious, or some other justification, “Oh please. Our kids turned out fine.” That’s not how biology or science works.

Much of these worlds is a matter of probability. Following best practice doesn’t mean guaranteeing your child will be fine. It doesn’t mean what old parents did was terrible while what we new parents do is great. It means new parents are (hopefully) an incremental increase in the probability children will be safe. A parent saying their kids turned out fine is meaningless reasoning. The odds they turned out fine were not as good as nowadays, provided science is moving in the right direction.

While we have halts and reversals, on balance science makes our practices better. It’s not the most comforting thing for a parent to realize they shouldn’t have been placing their child on their stomach to sleep, but that’s how knowledge works. I have no doubt when my kids have kids we’ll find out we shouldn’t have bounced them off walls like we did. Who would have predicted?

Some of the backlash over this and other alcohol studies is no doubt from moms who drank during their pregnancy. (If not dads who were accomplices.) As well as a group of moms who feels society constantly guilts them for their lack of perfection. Nobody is saying you’re a bad parent. Part of parenting is realizing how faulty you can be. But science says you shouldn’t have taken that risk, and there may be consequences even if they’re not immediately visible. Feelings, beliefs, are irrelevant.

Nobody coherent disputes being a woman is in various ways harder, that we have an egregious history when it comes to how women have been treated, but asserting a study done in part by nine women somehow has a guilt-all-women motive is pretty farfetched. It’s no different than women struggling to have the upper body strength of males. Where certain difficulties in being female will always be due to biology, not society.

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