Aug

2

A news article from yesterday's WSJ reports on a new study from Harvard that purports to demonstrate that breastfeeding enhances a baby's IQ by about 4 points, and that the effect has been isolated from other confounding variables. A pet peeve of mine is that as usual I don't have access to the original publication, which covered research that very likely received some form of funding via my tax dollars. Anyway, just from the news account, I am skeptical.

The study "followed 1,312 babies and mothers from 1999 to 2010..and then tested the children's intelligence at ages 3 and 7"

OK. Well I read elsewhere that:

"A large number of studies have suggested that low [omega fatty acids DHA and ARA] might be associated with problems with intelligence, vision, and behavior. Children fed standard formulas may have IQ's 5-9 points lower than breast-fed babies, even after correcting for other factors."

and that:

"until 2002, [omega fatty acid additives including DHA] were not present in the infant formulas available in the United States".

The new Harvard study covered 1999-2010, but the kids were assessed at age 7, so I assume that means that the kids' birth dates ranged only from 1999 to 2003. Therefore, essentially all of the formula-fed babies in the study were getting formula without the DHA and other additives that were added in 2002. So shouldn't I expect their IQs to be "5-9 points lower", simply because they were using pre-2002 formula? If so, that renders the study irrelevant and unnecessarily alarming to parents today who aren't able to breastfeed for whatever reason, and who are using today's omega-enhanced formulas (which is the only thing that is available anyway).

If that objection is somehow mistaken, then I have further skepticism about how ~1000 cases could be an adequate sample size, given what they need to show.

If they look at IQ vs only a single independent variable — breastfeeding — then it's pretty clear that their "N" is adequate. On a "back of the envelope" basis, I think they're claiming something like a 13% correlation between breastfeeding (measured on a continuum from no breastfeeding it all to >1 year of exclusive breastfeeding) and IQ, and I believe that a sample size of just a few hundred would be adequate. The measured correlation have something like 1/sqrt(N) ~ 3% standard error, much smaller than 13%.

However, they claim to be able to isolate the effect of breastfeeding from other confounding variables. The biggie is the mother's own IQ, which is highly correlated (50%?) with the baby's IQ, but also with breastfeeding, since smart moms on average are more likely to breastfeed. I hope that some statistics experts can help on this, but I do have the impression that it's notoriously difficult to isolate the effects of independent variables when they're strongly correlated with each other. Again, it would help if the original paper were accessible to the public, who probably paid for part or all of the research and have a real practical need to understand the results.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search