breastfeeding rocksA couple weeks ago I came across an academic article[1] stating that the benefits of breastfeeding were overstated using discordant sibling data from a national survey to ‘prove’ the point.  Now, this hadn’t hit the news yet, I took some time to digest the findings and the limitations and figured I’d have time to write it up slowly.  Ha!  Yesterday the proverbial shit hit the fan and news sites were sharing articles claiming things like the benefits of breastmilk are overrated or that breastmilk is “no better” for a baby than formula.  As I cringed over the interpretation, I figured it was time to throw my hat in after all…

What does the research say? (And, is it good research?)

The research used data from the children of the 1979 cohort of the National Longitudinal Survey of Youth which means the questions are standardized, but they reach a large number of kids which is great for researchers.  They analyzed over 8000 children, just under 1800 of which were what was termed ‘discordant’ sibling pairs, meaning that one of them was “breastfed” while the other was strictly formula-fed.  The authors then looked at the outcomes as assessed between 4 and 14 years of age on 11 specific outcomes: Three health outcomes (BMI, obesity, and asthma), three behavioural outcomes (hyperactivity, parental attachment, and behavioural compliance), and five academic achievement outcomes (Peabody Vocabulary Test, Peabody Reading Recognition, Peabody Math Ability, Weschler Intelligence Scale, and ‘scholastic competence’).  Notably, BMI, obesity, and the standardized achievement outcomes (everything except ‘scholastic competence’) were objectively measured, asthma and the behavioural outcomes were maternal report, and scholastic competence was child report and consisted six questions to assess the child’s sense of self-competence in school.

They ran separate models for the entire sample and the sibling only samples.  They also ran between-family estimates and within-family estimates, and for all models the usual covariates were included.  In short, the statistics are sound.  For the between-family estimates, there was a significant effect of breastfeeding on all of the variables and this remained when looking at sibling samples only.  However, when the within-family estimates were used, these results all but disappeared (with the exception of behavioural compliance which was weakened but still trending significance).  Within-family estimates are different in that they include the family as a variable meaning that any unobserved, but stable influence of the family is included.  These results held for breastfeeding ever (a yes/no dichotomy) and breastfeeding duration (in weeks).  Interestingly, the authors enjoy speaking about the null findings in the direction against breastfeeding as being meaningful and worthy of discussion (i.e., mother reported asthma) while stating that the null findings in the direction for breastfeeding are simply that: null.

So how good is the research?  Well, we’ll get into what’s missing next, but the first sign of something being off is where this is published.  A huge study, if it was conducted well, statistically robust, and had the right measurement of variables would be in something like the New England Journal of Medicine or Pediatrics (though there have been some real doozies in that one recently when trying to publish more social sciences instead of medicine) or Science or Nature.  It’s not.  It’s not in a bad journal by any means, but it raises the question of what area was limited that these other journals would have turned it down (because it only makes sense to try and go to one of the big ones first if you have groundbreaking data).  With that, let’s look at…

What is the research missing?

There are enough important variables missing that I can see why it wasn’t in a better journal.  First and foremost, the definition of breastfeeding is hardly informative.  The yes/no measure is almost useless as it would include mothers who breastfed once which we know is not indicative of any benefits, especially when looking long-term.  But even the duration in weeks (weeks!) is misleading because it doesn’t deal with the difference between exclusive and mixed breastfeeding.  Anyone who looks at the data on health and other benefits of breastfeeding is well aware that the strongest influences are for long-term, exclusive breastfeeding (which also happens to be the biological and historical normal).  The authors also fail to include what the average duration was so we have no idea if we’re looking at 3 weeks, 20 weeks, 5 weeks, etc. and the number of families at the higher ends.  If no families breastfed for any significantly long duration, there would be no way to find the proper relationships because you have range restriction on the key variable of interest.

Second, there is no information on the influence of these within-family effects, not even the actual weights in the regression equations.  That is, we have no idea why one child was breastfed but another not, what birth order it was, or what the interactions were with these influences.  We don’t know if the discordant siblings were mixed-fed and formula-fed which might have very different implications than exclusively-breastfed versus formula-fed.  Without this information, we’re missing the crucial part of the picture.  Notably, looking at the mean data they do provide, the discordant sibling sample had a much lower SES than those who did not have discordant siblings (i.e., the siblings were both either breastfed or not breastfed).  Discordant siblings were also more likely to have a child that was preterm, and was also less likely to have drank during pregnancy.  We just don’t know how these variables interact with breastfeeding which seems a crucial distinction.

Third, it’s missing a lot of child health outcomes relevant to the breastfeeding discussion.  I realize they were limited in outcome selection because it was based on national data, but the outcomes selected are not a lot of the standard ones associated with long-term breastfeeding outcomes which predominantly focus on health.  The WHO’s recent meta-analysis on the same topic[2] looked at cholesterol, blood pressure, obesity, type-2 diabetes, and intelligence, finding a tiny effect of breastfeeding duration on obesity (which was one of the measures here) and a small effect on intelligence (using some of the same measures included in this review).  The other three health outcomes – which included significant effects of breastfeeding on blood pressure and type-2 diabetes – were not included here.  Of course no other health outcomes, like cancer, were included which adds to this limitation.  In short, concluding that the benefits of breastfeeding are overstated or non-existent when only looking at a small subset of outcomes is a little disingenuous.

Fourth, the data is entirely missing the outcomes for mothers.  The rates of cancer for mothers is significantly lower long-term for women who breastfeed versus those who do not[3].  Even if there truly were no differences between breastfeeding and formula for babies (which is not the case, as you’ll see if you read the next paragraph), it would still benefit everyone for mothers to breastfeed as the alternative puts them at higher risk of dying.  According to recent research, the United States are losing 72,000 women prematurely because of a failure to reach “optimal breastfeeding rates” (defined as 90% of women breastfeeding exclusively for 6 months)[4].  Does that not matter?

Finally, and perhaps most importantly for the idiotic headlines we’re seeing, the data here speaks to long-term outcomes with no discussion or consideration of the immediate or short-term outcomes which strongly favour breastfeeding.  The authors of this report seem to just ignore that evidence in favour of focusing all discussion exclusively on long-term outcomes.  In the US alone, the immediate (< 1 year) cost of not reaching “optimal breastfeeding rates” is resulting in 911 deaths per year (almost all infants under 1 year) and a cost to the health care system of $13 billion a year[5].  I dare anyone to read that and suggest that breastfeeding is “no better” than formula.

What is the take-home message?

First I have to give credit to the authors for a section of their paper that has received almost zero attention: At the end they argue that instead of focusing our efforts on simply harassing women to breastfeed, we need to change the structure of our society.  That if we want to see improvements in long-term outcomes, we have to focus on the between-family differences by providing maternity leave, better health care, and a host of things relating to poverty or SES.   Myself, I think that we should also start looking at other possibilities for the within-family differences, such as style of parenting, which often remains similar across children (though definitely not always).  I do, however, laud their efforts to highlight problems in the current system instead of just focusing on the idea that formula is just as good, though sadly the media has ignored this part in favour of the “just use formula” aspect of the research.

Based on the research though, I think we can conclude that there are outcomes in which breastfeeding will not have an influence.  I have no idea if all of these outcomes here will fit this bill without knowing more about the duration and exclusivity of breastfeeding in this sample (though notably many of these outcomes have already been found to be low on the breastfeeding-effects totem pole so it’s not too surprising there weren’t significant differences here).  The idea that one can say breastfeeding is no better than formula or that the benefits are exaggerated is not supported by this data.  I do like, however, the use of discordant siblings and hope to see more of that research going forward as it may provide us with valuable information.

What the research does support is the idea that breastfeeding is not a magical elixir that will fix everything forever more.  Sadly I know there are people that treat it this way and it’s a problem as any evidence like this is then held up as proof that breastfeeding isn’t a panacea.  Breastfeeding is wonderful, natural, and in some areas of the world, it absolutely is the difference between life and death.  In our Western world, however, to assume breastfeeding has such long-term impacts that it would outweigh years of experiences ignores not only the wealth of other factors that influence these outcomes, but common sense.  Most people are aware of this and don’t try to make breastfeeding out to be more than it is.  It’s just that what it is, is still pretty darn cool.

Take the measures of intelligence, for example.  These authors and many others look at Western constructs of intelligence – namely verbal and mathematical IQ assessed using standardized tests – and although these are valuable, they are largely influenced by education and exposure to certain problems and bits of information and do not reflect unlearned intelligence.  As such, SES, education views in the household, school systems, parental support on homework and studying, etc. are all going to influence this far more than breastfeeding.  Even looking at health outcomes, how could one expect breastfeeding in infancy (especially short-term breastfeeding or mixed feeding that is predominantly formula) to overcome a steady diet of processed, non-nutritious foods for years after?  Now, if all of this were taken into account, might we see a difference then?  I don’t know, but I think it would be much more likely than when we don’t consider the other influences at all.

When we look at this research in conjunction with the wealth of other findings out there, we can see that the long-term effects of breastfeeding are small and perhaps even negligible for certain families where other factors are far more influential.  However, even if we took this data at face value and said it was the perfect study, why-oh-why would we only consider long-term outcomes in the assessment of feeding methods for infants in making statements about the overall efficacy of breastfeeding?  Since when did we not care at all about short-term consequences, especially when they include death that is significant at a population level?  Since when did we not care about the outcomes for mothers?  Or does our society only want to care about our influence on the economy, meaning they need to get us back to work and away from our kids as quickly as possible, even if it kills us sooner?

This study far from “proves” anything.  The journalists that have reported on it have once again taken liberties in their reporting that fall squarely on the side of irresponsible (however the authors themselves need to take some blame too here for blatantly ignoring a wealth of research that speaks to the “benefits” of breastfeeding short-term and for mothers, and for not speaking of their own limitations in this regard in their piece).  If you thought breastfeeding was so magical it would overcome anything, then this study should hopefully show you that no it’s not.  If you have a more realistic view of breastfeeding, I’m not sure this adds much to the discussion as you already know breastfeeding does provide things formula does not, but that in our society women use formula for many reasons.

My biggest fear with articles like that is that it completely detracts from the push to make breastfeeding easier for those who want to do it (i.e., those fighting for greater support and leave and all the other things necessary for many women to have a healthy, happy breastfeeding relationship) and to make breast milk more readily available for women who don’t want to turn to formula but cannot breastfeed for whatever reason.  I can only hope that flawed science doesn’t take us down a rabbit hole we have to dig our way out of yet again…



[1] Colen CG, Ramey DM.  Is breast truly best? Estimating the effects of breastfeeding on long-term child health and well-being in the United States using sibling comparisons.  Social Science & Medicine 2014. (Ahead of print)

[3] For a review with all the references, see: http://evolutionaryparenting.com/two-bs-and-the-big-c/

[4] Bartick M, Stuebe A, Schwarz EB, Luongo C, Reinhold AG, Foster EM.  Cost analysis of maternal disease associated with suboptimal breastfeeding.  Obstetrics & Gynecology 2013; 122: 111-9.

[5] Bartick M, Reinhold A.  The burden of suboptimal breastfeeding in the United States: a pediatric cost analysis.  Pediatrics 2010; 125: e1048-e1056.