Oh, really? Hanna Rosin in the Atlantic makes the case against breastfeeding

Hanna Rosin's new article in the Atlantic follows a script that one commonly sees in reporting about nutrition science.

We have the headline boldly claiming that everything you thought you knew is overturned ...

The Case Against Breast-Feeding

... and the mild claim, buried in the middle, that defeats only a straw dummy (formula is not as menacing as smoking).

So overall, yes, breast is probably best. But not so much better that formula deserves the label of “public health menace,” alongside smoking.

We have the belittling of important results, whose main defect seems to be disagreement with the author's story line.

Kramer followed 17,000 infants born in Belarus throughout their childhoods. He came up with a clever way to randomize his study, at least somewhat, without doing anything unethical. He took mothers who had already started nursing, and then subjected half of them to an intervention strongly encouraging them to nurse exclusively for several months. The intervention worked: many women nursed longer as a result. And extended breast-feeding did reduce the risk of a gastrointestinal infection by 40 percent. This result seems to be consistent with the protection that sIgA [an element of breastmilk] provides; in real life, it adds up to about four out of 100 babies having one less incident of diarrhea or vomiting.

And I think we may have an instructive misunderstanding or misquoting of some statistical results.

Rosin quotes one fascinating study by Evenhouse and Reilly that compares cross-sectional analysis (showing how different breastfeeding practices for children in different families are related to different outcomes) to a within-family fixed effects analysis (showing how different breastfeeding practices for children in the same family are related to different outcomes). The fixed effects analysis has an important advantage, by controlling for confounding characteristics of the family. The fixed effects analysis also has an important disadvantage, lower precision (much bigger standard errors) because there are fewer mothers who change breastfeeding practices for different siblings.

Because of the lower precision (bigger standard errors), the same estimate of the effect of breastfeeding -- such as the finding that breastfed children are lighter or brighter -- could be "statistically significant" in the cross-sectional analysis and "statistically insignificant" in the fixed effects analysis. The important thing to do, in such cases, is to look at the actual estimates to see if they are much different in the two analyses.

If I am understanding Evenhouse and Reilly's analysis correctly, the cross-sectional estimates showed the usual benefits of breastfeeding for many outcomes. The fixed effects estimates agreed fairly closely on most of these outcomes, but were statistically insignificant because of larger standard errors.

For one key cognitive function score (PVT score), even the fixed effects analysis found a benefit of breastfeeding. So, Evenhouse and Reilly, in their summary, have good things to say about breastfeeding.

The significant correlation between breastfeeding and PVT score in our within-family model provides more credible evidence of a causal link between breastfeeding and cognitive ability than do existing nonexperimental studies. The effect is large enough to matter, and it is lasting, persisting into adolescence. Stronger evidence of causality may argue for intensifying breastfeeding promotion, particularly among groups that suffer from high rates of academic failure and other problems that some researchers have correlated with lower IQ (e.g., incarceration, poverty, or welfare recipiency). Some of the same social problems that justify additional expenditures on education and Head Start, for example, may also warrant additional efforts to raise breastfeeding rates.

Our results also suggest, however, that many of the other long-term effects of breastfeeding have been overstated.

How does Rosin describe what these authors say?

Almost all the differences turned out to be statistically insignificant. For the most part, the “long-term effects of breast feeding have been overstated,” they wrote.

Notice how Rosin quotes selectively from just part of Evenhouse and Reilly's last sentence. The rest would disagree with her story line.