Breastfeeding is one of those weird subjects where people's emotional investment appears to be dramatically out of proportion with the measurable impact.[1] Women everywhere are shamed and derided if they choose to formula feed their children, from a wide variety of actors.
Personally, as a father, after breastfeeding one child and formula feeding another, I vastly prefer formula. It's orders of magnitude more convenient and proportionally less stressful. Especially if the mom is working and has to handle the responsibility of pumping on a regular schedule.
We took one of those "get ready for childbirth" classes and one subject they covered was breastfeeding. I'm still mad that the instructor tried to claim that breastfeeding was the more convenient option. Nothing is more convenient than being able to hand the baby to any other person and ask them to feed it for you!
[1] Another subject that shares this distinction is circumcision.
Yes, totally true - I should've mentioned these practical aspects too, which have to be weighed up against the (highly uncertain) benefits claimed in the studies.
Your response that formula is less stressful and more convenient has the same issue as the advice of birthing class instructor that you bemoan. It’s relative. I’m currently exclusively breastfeeding and my impression is that it’s very easy (likely because my baby latches well, I have no pain, I enjoy it, I don’t work, it soothes baby quickly, etc). People are different. That said, I do get a kind of off-putting feeling when I hear about moms who didn’t even try to breastfeed. I don’t know why I feel this way.
Agreed. I breastfed all four of my kids, and once past the neonatal days (which are stressful no matter how the baby is fed), found breastfeeding to be very stress free. One never has to worry about cleaning bottles, buying formula, heating milk, packing a full diaper bag of supplies to leave the house. Night-time feedings are also relatively easy, and then there's the stress-reducing benefits of prolactin and oxytocin! I would agree with Tim that if pumping and employment-related separation are involved, there is an added level of difficulty and stress. Sadly, the U.S. doesn't have many parent-friendly workplaces or parental leave policies that allow mothers to physically recover from childbirth and effectively establish breastfeeding in the early months.
P.S. Since feeding at the breast is the default means of feeding human beings, why is the discussion framed as if breastfeeding is the extraordinary intervention that has (or doesn't have) benefits, rather than seeing it as the biological norm and thus speaking of the possible negative effects of the actual intervention, which is formula feeding?
I disagree. My wife breastfed all of our children.
The hardest part for her was in the early weeks (latching on, mastitis) but afterwards it was just incredibly convenient. It actually made her a lot more autonomous as she could just feed our child whenever, wherever.
I think a lot of the discourse in the US is coloured by the lack of statutory maternity leave you get in Europe.
Tremendous and extremely valuable read Stuart! I really appreciate all the hard work that’s been put into this.
Just in case you missed it there is a typo in the text when you first talk about genetics (second bulletpoint):
“These confounding factors aren’t mutually-exclusive, by the way: the genetic effect could be operating as well as the environmental ones, and the. The”
RE publication bias - it should be possible to sell a negative result as a ‘hey, good news: you’re not harming your child’s IQ if you feed them formula’; given that many people can’t breastfeed, a ‘however you feed your kids is fine’ result seems to help more people than ‘breastfeeding increases IQ’.
My review of this meta-review of meta-reviews on breastfeeding:
1. It felt like you're a bit too dismissive/hand-wavey about the Belarussian/Brazilian studies, rather than trying to genuinely deconstruct them. I agree that they're a bit shakey as well, but the tone in which you write about them feels a bit like that's the weakest part of your post.
2. Do we necessarily expect research to exist that would show a *negative* correlation between breastfeeding and IQ, so that the funnel plot has some negative datapoints too? Are there no funnel plots where there's just genuinely no negative datapoints to report? I don't know the answer myself, but this question is why I wasn't entirely convinced by your response to the meta-paper.
3. Your response to the meta-paper and the authors response back are both behind a paywall, I'd like to the actual PDFs if you want to give readers a chance to look over them.
4. Is it possible that breastfeeding might improve IQ via better bonding, which in turn increases the amount of time spent by parents with their kids? Did any studies try to ask mothers how much *time* they've been spending with their kids and control for that?
5. As a corollary, perhaps the encouragement to breastfeed is basically an encouragement to spend more time with your kid? In other words, the official recommendation might not be about nutrition at all. I'd be curious to learn your thoughts on the internal reasoning of the experts who try to promote breastfeeding.
1. I don't know what else there is to say about the RCT other than digging in to note the difference in the better-blinded subsection (and I point to a letter further criticisms too). And at the end I say the Brazilian study has made me update back towards breastfeeding having an effect!
2. Even if there's a real effect you'd expect a few studies to find spuriously negative effects, just through chance, so the funnel plot should be symmmetrical (in theory anyway. You can get all sorts of weird-shaped funnel plots becuase of hidden selection biases in the studies).
3. Sci-Hub!
4. Yes - all the studies in the meta-analysis controlled for that.
5. In most of the papers I've seen, there's an (often seemingly quite half-hearted) ending section that discusses mechanism, and they bring in things like LC-PUFAs and so on. The impression I get is that they think it's mainly a nutritional, rather than social/interactive effect, though I suppose they might say it's a whole constellation of factors.
Just saw this after writing my other comment below (about causal mechanism)...
Just curious, how would you disaggregate the effect of bonding (or skin to skin contact) from the nutritional effects of breastfeeding to control for the potential confounding? I'm skeptical.
If the funnel plot shows no evidence of publication bias, it wouldn't necessarily include studies with negative correlations. It's more like we'd expect the size of the breastfeeding effect to jump around more in the studies with smaller samples, whereas we'd see more consistent effect sizes in the studies with larger samples.
Re: your 5th point, if a good RCT did ultimately establish a causal link, I think you're right to be curious how much the effect would be driven by social mechanisms. If there were established benefits, we wouldn't want to assume the mechanism is strictly nutritional. (But to Stuart's point, we can't be confident there's *any* effect now, so it's too soon to investigate possible mechanisms.)
I've read this piece a few times in the last year - I keep coming back to it because having two young children (who are both breastfed), I keep having to explain to people who don't understand research methods why the evidence for this is so weak. Now I just link them here. This article is a great service to humanity!
Reading the comments, I see the repeated difficulty of people saying "but there weren't any studies that said it was negative, so we should figure there is likely something positive, or at the least neutral." That is what the publication bias (and even some of the observational bias) is about. People want the "Yay! Breastfeeding!" answer to be true, so any contrary information just gradually goes away. The unpromising study gets abandoned and no further looking is done - or the study doesn't get published and just slowly gets forgotten. It only looks like there are no negative results. If that sounds far-fetched, know that this has been observed and demonstrated many times, especially after scientific thinking changes on a subject and everyone goes back and says "We really should have paid more attention to this twenty years ago."
We were very pro-breastfeeding around here when our sons were born 40 years ago. I still think the bonding, even the romance of it could be valuable. But one of the advantages of studying IQ is that it's more measurable, like height, or speed, than something like bonding or social adjustment.
If this research goes like the other attempts to show an environmental influence on IQ, the closer we get to a null hypothesis, the more people will claim that breastfeeding still improves intelligence, but the tests just aren't good enough to pick that up.
@ Nikita - I didn't find it dismissive at all. An alternate hypothesis is that you want one answer to be true. I don't know if that's true, only you can judge. But you should check.
"I’m not arguing against breastfeeding, by any means." So why not? Answer that and you've answered who there are so many poorly designed studies making the contrary argument.
Amazing post! One important confounding variables that is almost never included in breastfeeding research is the development of insufficient feeding complications like malnutrition, dehydration, jaundice and hypoglycemia from low breast milk supply. The assumption of breastfeeding research is that mother have roughly equivalent capacity to lactate and that breastfeeding vs. formula feeding is a matter of choice. We know from the history of infant feeding is that among infants who had to be bottle fed due to low or lack of breast milk availability, 30% died from a combination of malnutrition and ingestion of contaminated milk. In non-Western countries, like Brazil, lower availability or fewer social programs to provide safe breast milk substitutes to infants who need formula due to low milk supply may result in poorer health and future intelligence due to exposure to intermittent or chronic malnutrition. We know that the early insults of jaundice, dehydration and hypoglycemia can cause brain injury that result in permanent disability. More subtle cases of malnutrition or failure to thrive can similarly affect cognitive development, academic performance and future income earning potential. Early malnutrition have also been associated with future metabolic conditions like obesity, hypertension, and diabetes. The inequality in nutritional and health outcomes due to variations in breast milk supply in the population is yet another giant blind spot in breastfeeding research. It could easily be fixed by looking at medical records for evidence of these complications.
1) About the evolutionary reason for breastfeeding, I don't think that's such a strong argument. Breastfeeding evolved in the ancestral environment, where the alternative was not baby formula and not even pasteurized milk, but bush foods which are hard to impossible for a baby to digest. The WHO still recommends breastfeeding until 6 months, and in poor countries with poor sanitation that might very well be a good idea, and could even be supported by evidence (I haven't checked, but wouldn't be surprised if it checks out). That has far less relevance for our context.
2) Are there no studies which find (significant) negative effects of breastfeeding? Even by chance, if the effects are small, you'd expect some studies on some measures to fall there. If not, it's probably evidence of the effects being a little positive or, more likely, publication bias.
3) You've mentioned sibling studies extensively, but how about twin studies for the heritability of IQ? If those are credible in your eyes, then they live a lot less room for any shared environment factors, like breastfeeding, to affect IQ, especially at later ages. No?
4) I'm trying to quantify your conclusion that "we don't really know what the effects of breastfeeding on IQ are". And while I whole-heartedly support conducting more RCTs, I wanted to ask what is your current posterior distribution for the effect. Sounds like you're very surprised by +9 IQ points, which means that's in the tail of your distribution. So what does it roughly look like?
If there is an effect, especially one as big as the Brazilian studies suggest, then much of the gap in educational achievement in White and Black students in the US could be attributed to the 3x rate of breastfeeding amongst white mothers than black, 60%/20%. It would be great if the ultimate answer to these gaps wasn't racism or culture, but boobs.
There are plenty of natural experiments on breastfeeding and IQ since different (rich European) countries have vastly different cultures around breastfeeding. Despite these different breastfeeding levels there are no significant IQ differences in continental Western Europe. An indication as good as any that breastfeeding has at most a minuscule effect on cognitive ability.
There is another issue which is not addressed here, but nevertheless extremely important: IQ is an ordering metric -- it tells you where you stand relative to the wider population with some assumptions about a smooth distribution. This makes taking an average IQ somewhat problematic, since there is a huge difference between an intervention that causes 30 people to increase their IQ score from 100 to 101, and an observational study where the control group contained an additional person over test with an IQ of 70. Yet the effect on study outcome is the same. You would want to be extremely suspicious of any study where measured changes are dominated by outliers, since these kind of effects vastly increase noise and sensitivity to confounders. Maybe some of the studies address this, but I don't think I've ever seen a breastfeeding study (or report of one) where anybody questions whether mean IQ score is a valid thing to measure.
Breastfeeding is one of those weird subjects where people's emotional investment appears to be dramatically out of proportion with the measurable impact.[1] Women everywhere are shamed and derided if they choose to formula feed their children, from a wide variety of actors.
Personally, as a father, after breastfeeding one child and formula feeding another, I vastly prefer formula. It's orders of magnitude more convenient and proportionally less stressful. Especially if the mom is working and has to handle the responsibility of pumping on a regular schedule.
We took one of those "get ready for childbirth" classes and one subject they covered was breastfeeding. I'm still mad that the instructor tried to claim that breastfeeding was the more convenient option. Nothing is more convenient than being able to hand the baby to any other person and ask them to feed it for you!
[1] Another subject that shares this distinction is circumcision.
Yes, totally true - I should've mentioned these practical aspects too, which have to be weighed up against the (highly uncertain) benefits claimed in the studies.
Your response that formula is less stressful and more convenient has the same issue as the advice of birthing class instructor that you bemoan. It’s relative. I’m currently exclusively breastfeeding and my impression is that it’s very easy (likely because my baby latches well, I have no pain, I enjoy it, I don’t work, it soothes baby quickly, etc). People are different. That said, I do get a kind of off-putting feeling when I hear about moms who didn’t even try to breastfeed. I don’t know why I feel this way.
Agreed. I breastfed all four of my kids, and once past the neonatal days (which are stressful no matter how the baby is fed), found breastfeeding to be very stress free. One never has to worry about cleaning bottles, buying formula, heating milk, packing a full diaper bag of supplies to leave the house. Night-time feedings are also relatively easy, and then there's the stress-reducing benefits of prolactin and oxytocin! I would agree with Tim that if pumping and employment-related separation are involved, there is an added level of difficulty and stress. Sadly, the U.S. doesn't have many parent-friendly workplaces or parental leave policies that allow mothers to physically recover from childbirth and effectively establish breastfeeding in the early months.
P.S. Since feeding at the breast is the default means of feeding human beings, why is the discussion framed as if breastfeeding is the extraordinary intervention that has (or doesn't have) benefits, rather than seeing it as the biological norm and thus speaking of the possible negative effects of the actual intervention, which is formula feeding?
I disagree. My wife breastfed all of our children.
The hardest part for her was in the early weeks (latching on, mastitis) but afterwards it was just incredibly convenient. It actually made her a lot more autonomous as she could just feed our child whenever, wherever.
I think a lot of the discourse in the US is coloured by the lack of statutory maternity leave you get in Europe.
Lact evidence lol
Tremendous and extremely valuable read Stuart! I really appreciate all the hard work that’s been put into this.
Just in case you missed it there is a typo in the text when you first talk about genetics (second bulletpoint):
“These confounding factors aren’t mutually-exclusive, by the way: the genetic effect could be operating as well as the environmental ones, and the. The”
Oops! Now fixed. Thank you!
Great post!
RE publication bias - it should be possible to sell a negative result as a ‘hey, good news: you’re not harming your child’s IQ if you feed them formula’; given that many people can’t breastfeed, a ‘however you feed your kids is fine’ result seems to help more people than ‘breastfeeding increases IQ’.
My review of this meta-review of meta-reviews on breastfeeding:
1. It felt like you're a bit too dismissive/hand-wavey about the Belarussian/Brazilian studies, rather than trying to genuinely deconstruct them. I agree that they're a bit shakey as well, but the tone in which you write about them feels a bit like that's the weakest part of your post.
2. Do we necessarily expect research to exist that would show a *negative* correlation between breastfeeding and IQ, so that the funnel plot has some negative datapoints too? Are there no funnel plots where there's just genuinely no negative datapoints to report? I don't know the answer myself, but this question is why I wasn't entirely convinced by your response to the meta-paper.
3. Your response to the meta-paper and the authors response back are both behind a paywall, I'd like to the actual PDFs if you want to give readers a chance to look over them.
4. Is it possible that breastfeeding might improve IQ via better bonding, which in turn increases the amount of time spent by parents with their kids? Did any studies try to ask mothers how much *time* they've been spending with their kids and control for that?
5. As a corollary, perhaps the encouragement to breastfeed is basically an encouragement to spend more time with your kid? In other words, the official recommendation might not be about nutrition at all. I'd be curious to learn your thoughts on the internal reasoning of the experts who try to promote breastfeeding.
1. I don't know what else there is to say about the RCT other than digging in to note the difference in the better-blinded subsection (and I point to a letter further criticisms too). And at the end I say the Brazilian study has made me update back towards breastfeeding having an effect!
2. Even if there's a real effect you'd expect a few studies to find spuriously negative effects, just through chance, so the funnel plot should be symmmetrical (in theory anyway. You can get all sorts of weird-shaped funnel plots becuase of hidden selection biases in the studies).
3. Sci-Hub!
4. Yes - all the studies in the meta-analysis controlled for that.
5. In most of the papers I've seen, there's an (often seemingly quite half-hearted) ending section that discusses mechanism, and they bring in things like LC-PUFAs and so on. The impression I get is that they think it's mainly a nutritional, rather than social/interactive effect, though I suppose they might say it's a whole constellation of factors.
Just saw this after writing my other comment below (about causal mechanism)...
Just curious, how would you disaggregate the effect of bonding (or skin to skin contact) from the nutritional effects of breastfeeding to control for the potential confounding? I'm skeptical.
If the funnel plot shows no evidence of publication bias, it wouldn't necessarily include studies with negative correlations. It's more like we'd expect the size of the breastfeeding effect to jump around more in the studies with smaller samples, whereas we'd see more consistent effect sizes in the studies with larger samples.
Re: your 5th point, if a good RCT did ultimately establish a causal link, I think you're right to be curious how much the effect would be driven by social mechanisms. If there were established benefits, we wouldn't want to assume the mechanism is strictly nutritional. (But to Stuart's point, we can't be confident there's *any* effect now, so it's too soon to investigate possible mechanisms.)
I've read this piece a few times in the last year - I keep coming back to it because having two young children (who are both breastfed), I keep having to explain to people who don't understand research methods why the evidence for this is so weak. Now I just link them here. This article is a great service to humanity!
Reading the comments, I see the repeated difficulty of people saying "but there weren't any studies that said it was negative, so we should figure there is likely something positive, or at the least neutral." That is what the publication bias (and even some of the observational bias) is about. People want the "Yay! Breastfeeding!" answer to be true, so any contrary information just gradually goes away. The unpromising study gets abandoned and no further looking is done - or the study doesn't get published and just slowly gets forgotten. It only looks like there are no negative results. If that sounds far-fetched, know that this has been observed and demonstrated many times, especially after scientific thinking changes on a subject and everyone goes back and says "We really should have paid more attention to this twenty years ago."
We were very pro-breastfeeding around here when our sons were born 40 years ago. I still think the bonding, even the romance of it could be valuable. But one of the advantages of studying IQ is that it's more measurable, like height, or speed, than something like bonding or social adjustment.
If this research goes like the other attempts to show an environmental influence on IQ, the closer we get to a null hypothesis, the more people will claim that breastfeeding still improves intelligence, but the tests just aren't good enough to pick that up.
@ Nikita - I didn't find it dismissive at all. An alternate hypothesis is that you want one answer to be true. I don't know if that's true, only you can judge. But you should check.
Plenty of good reasons to breast feed a baby but to imagine that it will make your child smarter seems very low on the list of possible benefits.
"I’m not arguing against breastfeeding, by any means." So why not? Answer that and you've answered who there are so many poorly designed studies making the contrary argument.
Amazing post! One important confounding variables that is almost never included in breastfeeding research is the development of insufficient feeding complications like malnutrition, dehydration, jaundice and hypoglycemia from low breast milk supply. The assumption of breastfeeding research is that mother have roughly equivalent capacity to lactate and that breastfeeding vs. formula feeding is a matter of choice. We know from the history of infant feeding is that among infants who had to be bottle fed due to low or lack of breast milk availability, 30% died from a combination of malnutrition and ingestion of contaminated milk. In non-Western countries, like Brazil, lower availability or fewer social programs to provide safe breast milk substitutes to infants who need formula due to low milk supply may result in poorer health and future intelligence due to exposure to intermittent or chronic malnutrition. We know that the early insults of jaundice, dehydration and hypoglycemia can cause brain injury that result in permanent disability. More subtle cases of malnutrition or failure to thrive can similarly affect cognitive development, academic performance and future income earning potential. Early malnutrition have also been associated with future metabolic conditions like obesity, hypertension, and diabetes. The inequality in nutritional and health outcomes due to variations in breast milk supply in the population is yet another giant blind spot in breastfeeding research. It could easily be fixed by looking at medical records for evidence of these complications.
Excellent article!
A few comments and questions:
1) About the evolutionary reason for breastfeeding, I don't think that's such a strong argument. Breastfeeding evolved in the ancestral environment, where the alternative was not baby formula and not even pasteurized milk, but bush foods which are hard to impossible for a baby to digest. The WHO still recommends breastfeeding until 6 months, and in poor countries with poor sanitation that might very well be a good idea, and could even be supported by evidence (I haven't checked, but wouldn't be surprised if it checks out). That has far less relevance for our context.
2) Are there no studies which find (significant) negative effects of breastfeeding? Even by chance, if the effects are small, you'd expect some studies on some measures to fall there. If not, it's probably evidence of the effects being a little positive or, more likely, publication bias.
3) You've mentioned sibling studies extensively, but how about twin studies for the heritability of IQ? If those are credible in your eyes, then they live a lot less room for any shared environment factors, like breastfeeding, to affect IQ, especially at later ages. No?
4) I'm trying to quantify your conclusion that "we don't really know what the effects of breastfeeding on IQ are". And while I whole-heartedly support conducting more RCTs, I wanted to ask what is your current posterior distribution for the effect. Sounds like you're very surprised by +9 IQ points, which means that's in the tail of your distribution. So what does it roughly look like?
Thanks again!
If there is an effect, especially one as big as the Brazilian studies suggest, then much of the gap in educational achievement in White and Black students in the US could be attributed to the 3x rate of breastfeeding amongst white mothers than black, 60%/20%. It would be great if the ultimate answer to these gaps wasn't racism or culture, but boobs.
did a quick search of "twin", nothing came up...
This post spurred me to write kind of a reply: https://woodfromeden.substack.com/p/sweden-is-smarter-than-france-or
There are plenty of natural experiments on breastfeeding and IQ since different (rich European) countries have vastly different cultures around breastfeeding. Despite these different breastfeeding levels there are no significant IQ differences in continental Western Europe. An indication as good as any that breastfeeding has at most a minuscule effect on cognitive ability.
There is another issue which is not addressed here, but nevertheless extremely important: IQ is an ordering metric -- it tells you where you stand relative to the wider population with some assumptions about a smooth distribution. This makes taking an average IQ somewhat problematic, since there is a huge difference between an intervention that causes 30 people to increase their IQ score from 100 to 101, and an observational study where the control group contained an additional person over test with an IQ of 70. Yet the effect on study outcome is the same. You would want to be extremely suspicious of any study where measured changes are dominated by outliers, since these kind of effects vastly increase noise and sensitivity to confounders. Maybe some of the studies address this, but I don't think I've ever seen a breastfeeding study (or report of one) where anybody questions whether mean IQ score is a valid thing to measure.