SSC Journal Club: Childhood Trauma And Cognition

This month’s American Journal of Psychiatry includes Danese et al, Origins Of Cognitive Deficits In Victimized Children. Previous studies had found that abused children had lower IQ. They concluded that the severe stress of being abused must decrease brain function. Danese et al challenge that assumption in the context of the new paradigm of shared-environment-skeptical psychiatry.

They looked at some of the big developmental studies that had followed thousands of children from birth and evaluated them throughout the years. Most of these kids had gotten a bunch of different IQ tests every year, starting as young as 3. These children had also been screened for abuse, both by questions that tended to predict abuse (for example, asking parents whether they ever “punished their children harshly”) and by asking the kids, once they were grown up, whether they’d ever been abused or not.

The hope was to find children who hadn’t been abused before their first IQ test at age 3, but who started being abused before their last IQ test at age 18. Then they would see whether IQ at age 3 predicted IQ at age 18 in the same way it would in unabused children, or whether IQ was lower than predicted, presumably because of the abuse.

Before the adjustments, they found the same thing as every other study – abused children had lower IQ than unabused children (p < 0.01, B = 0.1):

Consistent with previous research, we found that adolescents and adults with a history of childhood victimization have pervasive deficits in clinically significant cognitive functions, including both general intelligence and more specific measures of executive unction, processing speed, memory, perceptual reasoning, and verbal comprehension.

After the adjustments, there was no difference between IQs in the two groups (p = 0.13, B = 0.03):

In contrast to the conventional causal interpretation of these findings, our longitudinal prospective design revealed that cognitive deficits in victimized adolescents and adults were largely explained by cognitive deficits present before the observational period for childhood victimization and by nonspecific effects of childhood socioeconomic disadvantage. On the one hand, the results are consistent with the high heritability of cognitive functions, their strong continuity across the life course, and the stable cognitive deficits previously described in children exposed to adversity. On the other hand, they are inconsistent with the causal effects of early-life stress on brain function reported in experimental animal models.

The study raises the possibility that kids from poorer families were more likely to get abused and to have lower IQs. Other possibilities include abusers (like criminals) having lower-than-average IQs and passing them on to their kids, or low-IQ kids being less able to figure out ways to escape abuse.

I find this interesting for two reasons.

First, twin studies famously find that early family experiences (the “shared environment”) are less important than previously believed. This has led to a small cottage industry of trying to deny twin studies and saying they don’t work. There’s a lot of interesting debate in this area, and by “interesting” I mean “annoying and interminable”. Maybe the best solution is to route around the whole problem by viewing twin studies as an incentive to conduct really rigorous non-twin studies to see if alternate methodologies can confirm the same surprising conclusions. Danese et al is a rare example of this being done right.

Second, a lot of discussion of the limited role of the shared environment adds the caveat “of course, we’re not talking about actual abuse here. Of course that affects outcomes.” It probably affects some of them. But here we see that it’s not necessarily true along every axis.

A recent twin study showed that child abuse doesn’t increase risk for ADHD, autism, or learning disorders. But I don’t think most people expected it to increase the risk of these disorders, which are traditionally viewed as really genetic.

A 2010 study found that it didn’t increase the risk of “conduct problems” broadly defined. These are traditionally viewed as linked with child abuse, so that’s interesting.

A 2013 study found that child abuse mostly didn’t explain personality disorders. Traditionally borderline personality disorder in particular is viewed as almost invariably associated with abuse, but if I’m reading the study right abuse really only explained about 1% of the variance in who had the condition. And although this called itself a “co-twin analysis”, it wasn’t a real twin study – it used non-abused twins as the control group for abused twins, but it didn’t check if they were identical or non-identical; this controls for family socioeconomic status but not for genetics. This leaves open the possibility that even the 1% remaining abuse-related variance might not be real.

The strongest study I could find was this one linking childhood sexual abuse to a threefold lifetime increase in psychiatric disorders. It, too, was a cotwin analysis and so not genetically controlled. And although I am not going to win any sensitivity points for saying that genetics can affect likelihood of being sexually abused, this is my impression of a lot of the cases I’ve seen – for example, children with a lot of of preexisting issues and bad relationships with their parents are more willing to secretly meet predators who have been grooming them online.

On the other hand, there are also reasons to believe these studies might understate the dangers of abuse. For example, the Danese et al study sampled 3000 kids who were supposed to be representative of the communities they came from. Suppose only 1% of children experience very severe child abuse. Then they would only have a sample of 30 – far too small to find any interesting conclusions. Maybe most of what they’re picking up is differences between not-abused-at-all kids and only-mildly-abused kids, and we shouldn’t be too surprised if they don’t find anything major.

I don’t want to underplay the role of child abuse. My impression from anecdotal evidence is that it does seem to have serious negative effects on kids later in life. For example, I’ve seen some kids from well-off, genetically-healthy families who were sexually abused by a coach or a neighbor or a priest or someone and who became nervous wrecks later in life, unable to stop thinking about it or to stop feeling guilty. Even good studies usually find that trauma is a risk factor for PTSD (thank goodness), proving that highly stressful events can be psychologically scarring and lead to future problems. Once you’ve accepted that, you’ve got most of the assumptions you need to reconstruct the case for child abuse causing psychological problems.

I think the safest assumption is that child abuse that rises to the level of trauma is associated with a PTSD-like picture (although I’ve heard some experts discourage thinking of it as literally PTSD) that seriously affects later-life outcomes. I think it’s less likely that child abuse has many effects that are wildly different from what we see in normal PTSD. And since normal PTSD doesn’t decrease IQ, I’m not surprised to hear that child abuse doesn’t do that either.