Does breastfeeding increase IQ or do breastfeeding complications decrease it?

Glass 3d buttons. Up and down

There’s a new paper on breastfeeding and IQ.

Is breast feeding associated with offspring IQ at age 5? Findings from prospective cohort: Lifestyle During Pregnancy Study yielded surprising results:

[pullquote align=”right” cite=”” link=”” color=”” class=”” size=””]We need to know not merely how long women breastfeed, but why they stop.[/pullquote]

In multivariable linear regression analyses adjusted for potential confounders breast feeding was associated with child IQ at 5 years (categorical χ2 test for overall association p=0.03). Compared with children who were breast fed ≤1 month, children breast fed for 2–3, 4–6, 7–9 and 10 or more months had 3.06 (95% CI 0.39 to 5.72), 2.03 (95% CI −0.38 to 4.44), 3.53 (95% CI 1.18 to 5.87) and 3.28 (95% CI 0.88 to 5.67) points higher IQ after adjustment for core confounders, respectively. There was no dose–response relation and further analyses indicated that the main difference in IQ was between breast feeding ≤1 month versus >1 month.

Here are the results of verbal IQ:


And performance IQ:


The authors concluded:

Breastfeeding duration of 1 month or shorter compared with longer periods was associated with approximately three points lower IQ, but there was no evidence of a dose–response relation in this prospective birth cohort, where we were able to adjust for some of the most critical confounders, including maternal intelligence.

That makes little sense on its face. If breastfeeding truly improves IQ, one would expect a dose response relationship with longer periods of breastfeeding leading to higher IQ. Moreover, one would expect little to no impact from very short periods of breastfeeding.

The authors chose to put a positive spin on a paradoxical result:

Our finding of a three point difference in IQ associated with any duration of breast feeding longer than 1 month is in support of current recommendations, and is even a relaxed message to mothers who struggle with exclusive breast feeding.

But that’s not the only conclusion you could draw. The same data could be used to argue that the babies who breastfed for less than a month were harmed by adverse effects of breastfeeding itself. Instead of increasing IQ, breastfeeding had no impact and breastfeeding complications actually decreased IQ. It’s not a possibility that the authors ever considered since nearly all breastfeeding research starts with the unfounded assumptions that breastfeeding must have benefits and couldn’t have harmful effects.

The study itself has some very real strengths but also some serious weaknesses.

It’s chief weakness is that it reflects secondary findings from a study designed to assess the impact of maternal alcohol intake.

The LPDS (Lifestyle During Pregnancy Study) consists of 3478 mother–child dyads selected from the DNBC with oversampling of pregnant women with moderate weekly alcohol intake, alcohol binge drinkers and women with high versus low fish intake, iron intake and duration of breast feeding, respectively.

Secondary findings are often the result of outcome switching, an issue with serious ramifications for the integrity and reproducibility of the research.

As John Ioannidis and colleagues explain:

Outcome switching refers to the possibility of changing the outcomes of interest in the study depending on the observed results. A researcher may include ten variables that could be considered outcomes of the research, and — once the results are known — intentionally or unintentionally select the subset of outcomes that show statistically significant results as the outcomes of interest. The consequence is an increase in the likelihood that reported results are spurious by leveraging chance, while negative evidence gets ignored. This is one of several related research practices that can inflate spurious findings when analysis decisions are made with knowledge of the observed data, such as selection of models, exclusion rules and covariates. Such data-contingent analysis decisions constitute what has become known as P-hacking …

There’s also the possibility that the variable that was originally studied (in this case alcohol intake) is a confounding factor in any study of the secondary variable (in this case IQ). The authors do acknowledge this problem and attempt to adjust for it.

A major strength of the study is that adjustment for critical confounding variables including, most importantly, maternal IQ. Most studies on breastfeeding and child IQ have failed to take maternal IQ into account, which renders their findings highly suspect.

Only 6 out of 1385 women in the study (0.4%) chose not to attempt breastfeeding.

In our study sample, we categorised the shortest duration as ≤1 month, since very few women reported breastfeeding duration shorter than this, reflecting that by far the majority of mothers in Denmark choose to breast feed their children.

The authors understand that those who don’t even try differ from other Danish women in important ways:

Adding to the difficulty of obtaining an exposure group with shorter duration of breast feeding is the fact, that women who from the beginning choose not to breast feed may be different from those who do breast feed; for example, women who rely on medication for various reasons may choose not to breast feed because of concerns that medication in the breastmilk may harm the infant…

So far, so good, but the authors fail to consider that those who stop breastfeeding after less than a month may also differ from other Danish woman in a critical way. Their infants may have suffered medical complications from breastfeeding like dehydration or jaundice from insufficient milk supply. It seems never to have occurred to the authors that breastfeeding can have risks as well as benefits. Since up to 15% of first time mothers may have insufficient breastmilk, particularly in the early days of birth, a substantial proportion of babies will likely suffer serious consequences of any effort to promote exclusive breastfeeding.

The data the authors provide suggest that may indeed be the case:

Women who breast fed for less than 1 month compared with 7–9 and more than 10 months were generally younger, they were more likely to be nulliparous (had not previously given birth), have higher BMI, to have been smokers during pregnancy or to have their children be exposed to tobacco smoke postnatally, and have lower IQ…

Nulliparity and higher BMI are both risk factors for insufficient breastmilk.

More notable was this:


Both maternal IQ and education were linearly associated with breastfeeding duration EXCEPT for duration less than a month. As the chart above shows, Maternal IQ and education were lowest for those who breastfed 2-3 months, whereas IQ and education for those who breastfed for less than a month were equal to the mean for the group. That’s just what you would expect if breastfeeding for less than a month were not a choice but a necessity due to medical factors.

The authors believe they found this:

We found no clear dose response relation of breastfeeding duration with child cognitive development in our data; rather, our results point to a difference in IQ of approximately three points between children who are breast fed for a short period of 1 month or less compared with those who are breast fed longer.

But they may have found the opposite: breastfeeding has no impact on IQ, but breastfeeding complications lead to a decrease in IQ.

There’s one way to find out. We need to know not merely how long women breastfeed, but why they stop. The IQ of children whose mothers chose not to breastfeed for personal — not medical — reasons may be no different than the IQ of children who were breastfed for more than 1 month. That would upend everything we believe we know about the benefits of breastfeeding.