Four Little Pigs
Three recent infant feeding trials' findings, shortcomings, and the political and social dimensions of the science; and a proposed target trial
Reader, it’s been a long time. Three months without a word. I love my work, hated the silence, and felt guilty — but can’t help diving deep or not at all. Between minor but pesky health problems (anemia, sinusitis), major life changes (moving, embracing new love), and the coal mines of parenting (potty-training and transitioning my toddler into daycare), it was all I could do to eat more chocolate and dream of the day when I could unpack the medical literature again. Thank you for your patience.
Once upon a time, there were three little pigs, each working in his own interests in response to perverse incentives. Their different psychosocial milieux shaped their construction of different houses — structures with scientific façades that were actually built on myths, rather than on scientific evidence. (The latter would require thinking about causes first.)
These houses, in turn, generated different responses from the big, bad wolves — the pigs’ natural enemies. These responses had little to do with the underlying evidence, and a lot to do with power. All three pigs had enough political and social power that their houses remained standing — as bad science usually does.
A better way is possible, if we only cross the road… It involves envisioning an idealized trial that could better avoid risk of bias when it comes time to analyze the evidence. Real trials are always imperfect; they always have flaws, and we should choose those flaws wisely.
The houses I want to discuss are actually infant feeding trials. These are randomized trials that compare different ways of feeding infants. The scientists running these studies incur criticism from people with different interests — conflicts of interest. The scientific discourse is bad about contextualizing science and science criticism in terms of its political and social stakes, with many audiences confusing rhetoric of neutrality with its reality; but we can do better science by saying out loud the quiet part of psychosocial entanglements.
The first trial, PRIMES, supplemented breastfeeding in sub-Saharan Africa with small amounts of formula. It didn’t establish either growth benefits or other harms, highlighting the inadequacy of minimal supplementation. Critics charged the trial was unethical, because it dared to contest the way of doing things their own research had promoted — a way associated with common and preventable harm to infants.
The second trial, MILK, pitted donor breastmilk versus formula for extremely preterm U.S. infants receiving minimal milk from their own mothers. It showed a possible survival benefit associated with formula. Critics ignored that evidence, trumpeting instead the importance of breastmilk. The researchers themselves prematurely halted the trial due to rising belief in neonatal centers that breast was best.
The third trial assessed the effects of a particular strain of the popular probiotic Lactobacillus reuteri on infant gastrointestinal problems and colic. Results confirmed L. reuteri’s established benefits. However, the researchers failed to publish all results preregistered as outcomes of interest. They stood to profit from selling an associated product, and no one had an interest in scrutinizing their science.
A fourth, fantasy infant feeding trial would be pessimistic about our ability to measure breastmilk amount, its nutritional and probiotic content, and common breastfeeding problems that might lead to accidental infant starvation and/or breastfeeding cessation. It would minimize the risk of bias from these sources by using a two-by-two factorial design to build on this literature by assessing how different infant feeding paradigms compare with and without probiotics, in terms of relatively easily measurable infant and child health and developmental outcomes like infections and delays.
Starting with causes, it would try to assess whether minimizing potential nutritional inadequacies in early life drives better outcomes. To do this, it would pit the “breast is best” or “exclusive breastfeeding” paradigm — which says to give babies only breastmilk for six months — against a “fed is best” alternative which encourages formula or other appropriate supplemental milk early, often, and enough. Again putting causes first, it would also ask whether probiotics contribute to outcome differences, since a wide array of evidence suggests they improve nutritional status.
On one hand, critics invested in the status quo may oppose such a trial as unethical because it rejects their ideology; this has long been exclusive breastfeeding ideologues’s argument against doing better science. On the other hand, it could also be vulnerable to criticisms that the evidence bases for prelacteal (early), supplementary (milk), and complementary (food) feeding and probiotics is already so strong that it is unethical to withhold those treatments from some subjects. Ultimately, current infant feeding guidelines incur substantial possible risks to newborns in exchange for no established benefits. Initiating the indicated paradigm shift may require evidence from such a trial.
Scene 1: Black babies at risk of malnourishment. Top hospitals for moms and infants in Guinea-Bissau and Uganda, sub-Saharan Africa. A house of straw called PRIMES.
The first little pig built a house out of exclusive breastfeeding (“breast is best”) ideology — the myth that having only breastmilk to eat or drink for the first six months of life is best for babies, even though near-universal breastfeeding insufficiencies cause common and preventable harm to newborns. Previously, evidence from the PROMISE-EBF trial had shown that exclusive breastfeeding promotion in sub-Saharan Africa successfully decreased prelacteal feeding in Burkina Faso — a decrease that correlated with an increase in neonatal deaths and the highest-ever measured perinatal mortality rate in Burkina Faso. These deaths were significantly predicted by nulliparity (first-time motherhood), consistent with a possible link with insufficient breastmilk (see again). Other apparent, accidental harms of the trial’s exclusive breastfeeding promotion included more wasting in Uganda, lower ponderal growth in Uganda and Burkina Faso, and a possible decrease in general cognition scores (95% CI -0.40 to 0.05). These harms came in exchange for no proven benefits in terms of morbidity or mortality, including the prevalence of diarrheal disease.
Some people might say that, if it looks like it harms and may kill babies, then you shouldn’t do it. That would mean demolishing the “breast is best” myth and prioritizing harm prevention in new infant feeding guidelines, instead. This is not in the interests of professionals who are invested in the status quo, like leading breastfeeding researchers. And it is not what they did.
In “Breastfeeding and Once-Daily Small-Volume Formula Supplementation to Prevent Infant Growth Impairment” (Pediatrics, Vol. 153, No. 1, Jan. 2024), Flaherman et al responded to PROMISE’s and other evidence of harm from exclusive breastfeeding promotion in sub-Saharan Africa by running the PRIMES trial (Preventing Infant Malnutrition with Early Supplementation). This was an effort to change the status quo a little bit, eventually, to harm babies less while still promoting the exclusive breastfeeding paradigm that causes common and preventable harm. Flaherman et al randomized 324 healthy, breastfeeding newborns in Guinea-Bissau and Uganda to receive one month of 59 mL of premixed, sterile, single-use formula bottles, to be given after a completed breastfeeding session. But, probably because the protocol was to offer formula after breastmilk, the babies only consumed, on average, around 30 mL of formula per day — 31.9 ± 11.8 mL, to be exact. For reference, the Department of Health in the Australian state of Victoria says newborns this age need 150 mL of prepared formula per kilogram of body weight, every day. (A can of Coke is 330 mL.)
This relatively small volume of formula consumed for this small duration of time (a month) “did not alter growth or cause breastfeeding cessation or AEs [adverse events].” PRIMES failed to prevent rampant growth impairment at six months “including stunting (27%), wasting (4%), and being underweight (14%).” As is typical, the babies in the treatment group got hungrier over the study month/period: “Among infants in the intervention group, the median (interquartile range [IQR]) daily volume of ESVFS feeding ranged from 14.79 (7.39–22.18) mL at 4 days of age to 36.97 (14.79–51.75) mL at 30 days of age…” Then, the trial — and its formula provision — stopped. Also as one would expect, the babies in the control group were probably hungrier than the babies in the treatment group, so their caretakers gave them more other supplementation in response to hunger cues, in the absence of study formula provision: “Nonformula supplementation, including milk, water, sweetened water, juice, tea, and other liquids, occurred in the first 30 days for 19 infants (12%) in the intervention group and 35 (21%) controls (P = .02).”
So PRIMES’s incrementalist offer was not enough. Enough what? contains several open empirical questions: Not enough supplemental milk? Not enough complementary feeding (the term of art for other baby food)? Not enough time (duration) providing access to hygienic, nutritionally appropriate supplementary calories? Not enough time between nursing sessions and supplementary feeds? Not enough flexibility in protocol — not enough listening to babies’ hunger cues and mothers’ experiences?
Offering supplementary milk early, often, and enough to babies at risk of common and preventable harm from insufficient breastmilk is a no-brainer. But researchers proposing research to figure out exactly what this means and how to do it in sub-Saharan African would probably have to bankroll it by going to big formula companies (i.e., multinational food conglomerates) and saying: “Sorry, we’ve been wrong for fifty years; turns out you’re not the Devil. Now please help us figure out how to feed these at-risk babies by putting some of your massive philanthropy budgets into serious research on this problem, but be good multinational conglomerates — and don’t try to get people who don’t need them hooked on your products. Give them quality, put the public interest ahead of your bottom line.”
This may sound far-fetched. But University of Dundee Honorary Professor of Molecular and Clinical Medicine Stewart Forsyth has argued such collaborations are essential to combatting malnutrition with science, proposed a framework for safeguards against corporate abuses, and envisioned its use in a new era of evidence-based infant feeding guidelines. However, it’s not clear Big Food could solve this problem by throwing money at it, either; it’s the shame of global public health that most of the world can’t get a glass of clean drinking water.
Alternately, leading Western researchers could (also) take seriously the fact that there have been local infant feeding traditions worldwide predating the late 1970s ascendence of the “breast is best” myth. They could go to the people in these countries, like Burkina Faso — which saw its highest-ever-recorded neonatal mortality rate during PROMISE — and say: “You were doing something else before we intervened, including prelacteal feeding. It looks like maybe it was better for babies than what we tried. So much better that we may have even accidentally killed some babies — sorry. What were you doing, and why?”
The logistical challenges of getting formula and clean water to babies who need it in these countries are serious. And the international regime around infant feeding is currently set up — for good reason — to constrain companies that could help. In particular, the WHO formula marketing Code bans formula donations. Discouraging formula donations to starving babies would not seem to be in the babies’ interests.
Of course, formula companies do have stakes in raising formula usage rates. But other actors (international NGOs, states, healthcare professionals) also have stakes — in defending the status quo. A status quo that causes common and preventable harm to infants.
By contrast, the people who have the most direct stakes in improving infant health here are locals whose babies’ health could be compromised by pursuit of both those stakes (i.e., those held by multinational corporations on one hand, and international researchers/NGOs/activists/status quo defenders on the other). Their voices have been largely omitted from infant feeding research like Flaherman et al’s, which continues to assume that the West knows best. But maybe local knowledge has something to say about what we don’t know we don’t know when we talk about infant feeding best practices in sub-Saharan Africa. If it does, that knowledge will likely be of increasing practical importance, as climate change worsens relevant conditions like extreme heat (which may contribute to breastfeeding insufficiencies) and heavy rains (which can contribute to water contamination problems).
We (producers of scientific knowledge) need to listen to people and respect their expertise instead of assuming we know best — a charge often lobbed at the international NGO aid world from global southern corners. Instead, scientists tend to think of valid scientific knowledge as coming from scientists, and not from ordinary people themselves — a cultural fantasy of escape from the socio-cultural embeddedness we cannot escape. Then, because there is a lot of human capital that goes into knowing how to play the scientific discourse game, it can be difficult for alternative perspectives like these to gain platforms — to become part of the scientific conversation.
So we might expect some relevant critical voices to be conspicuously absent from the scientific discourse here. And they are. But there are still critical voices to be found…
Enter: Wolf.
To some actors with stakes in defending the status quo, the PRIMES trial was unconscionable. And they put their two cents in the scientific record before the trial had even concluded. In “Questioning the ethics of international research on formula milk supplementation in low-income African countries: response,” Doherty et al (BMJ Glob Health, Vol. 7, No. 9, 2022) spout exclusive breastfeeding myths to question the ethics of Flaherman et al’s study. The trial, they contend, “violates basic ethical principles and human rights,” because “There is overwhelming evidence of the benefits of breastfeeding, especially EBF, for mothers, children and societies.” There is no such evidence, but it has long been normal for infant feeding literature to contain such statements.
Doherty et al go on to call it a myth that “that formula milk is necessary and that breast milk is inadequate for infant nutrition” — defending current infant feeding recommendations which themselves lack sufficient evidentiary basis, and ignoring current norms that do acknowledge breastmilk’s inadequacy for infant nutrition (e.g., lacking vitamin D). Doherty et al also argue the intervention wouldn’t scale up — implicitly defending the current legal and ethical regime that indeed precludes it from scaling up. And, in considering issues of beneficence (maximizing benefits to study participants and minimizing risks), informed consent, and justice, Doherty et al ignore the possibility of common and preventable harm to newborns from the very exclusive breastfeeding practices they themselves promote and defend.
This attack serves Doherty et al’s interests by defending the status quo against researchers whose work (very weakly) threatened it. For context, three of the criticism’s authors (Ingunn M.S. Engebretsen, Tanya Doherty, Thorkild Tylleskä) were part of the PROMISE team. This criticism is part of a longstanding argument between the teams, with this particular response itself responding to a response to a commentary critical of the published study protocol, on which Doherty was the editor. Doherty and two colleagues also commented on the article online, making some good points (e.g., the Flaherman et al should have presented data on adverse events by study arm according to the CONSORT checklist) along with some richly ironic ones (e.g., Flaherman et al reported receiving funding from companies like Enfamil, a possible conflict of interest).
In other words, Doherty et al include some of the same researchers whose own previous work promoting exclusive breastfeeding in these countries (PROMISE) appears to have caused preventable harm to newborns, including deaths. Liability and other interests preclude these scientists admitting that. (“Just because it looks like we killed some babies, doesn’t mean we should stop running trials or promoting our intervention!”) Power being what it is, liability concerns stemming from killing babies in sub-Saharan Africa are probably pretty small. The affected country settings tend to be too poor to develop good cause of death data in the aggregate, much less support infrastructure to let mothers sue foreign researchers for individual wrongful infant deaths.
Still, the incentives at work here are clear: Researchers who have invested their careers promoting myth-based exclusive breastfeeding practices risk dying social and professional deaths if they admit that they may have been wrong, their work may cause common and preventable harm (particularly among some of the most vulnerable babies in the world, and including deaths), and current infant feeding recommendations must change to prioritize harm prevention. So instead, confronted by a professional movement that implicitly charges that the status quo may indeed be harmful, they go on the offensive. They have a lot to lose, and a lot of power with which to defend their interests.
Why, then, did the first little pig’s house remain standing? Why didn’t Doherty et al stop Flaherman et al from running their avowedly unethical trial, publishing the results, or keeping the publication in the scientific record?
The opponents are well-matched in terms of social and political power. In addition to being about equally well-positioned in the field (both sides include top breastfeeding researchers), they’re part of a larger discourse in which the dominant “breast is best” narrative is increasingly contested, including by families and healthcare practitioners who’ve seen their children and patients preventably harmed by myth-based infant feeding practices. Liability for these harms may yet accrue to particular healthcare practitioners and researchers in some cases.
For instance, citing the results of records released under the Freedom of Information Act, I have previously criticized Flaherman for conducting research that preventably harmed newborns in the name of exclusive breastfeeding. Someone should follow up to see if the newborns rehospitalized for starvation complications in Flaherman’s previous early limited formula supplementation trials went on to have neurodevelopmental problems associated with those complications. But the people with access to the relevant data have incentives to not check; they’re liable.
However, social and political dominance of the exclusive breastfeeding ideology means that researchers, healthcare practitioners, and parents who accidentally harm infants by implementing current infant feeding guidelines are unlikely to be held accountable. Conversely, people who contest that paradigm are more vulnerable. While it may look like a disappointing flop, PRIMES may well have been designed to buffer against that asymmetrical vulnerability by generating exactly the evidence that it generated: no harm and no benefit. “Looks like we have to rejigger the status quo just a tiny bit more,” Flaherman et al can argue in a follow-up. From their socially and politically embedded perspective, the first burden of proof was on researchers contesting the dominant paradigm to establish that a very small, short-term formula supplementation treatment didn’t appear to do harm, and they did that.
The harm they were worried about was not just in terms of breastfeeding cessation, but also in terms of necrotizing enterocolitis (NEC), says PRIMES Trial Study Coordinator Joan Murungi (May 5, 2023 email correspondence with Ac. Prof. Dr. José Luis Diaz Rossello, Docente Libre Honorario, Pediatra Perinatal, Maternidad Universitaria, Universidad de la República, Uruguay). NEC is a rare, horrific, and potentially lethal intestinal inflammation and infection to which premies are especially vulnerable. Full-term infants, however, are vulnerable, too, with about ten percent of NEC cases seen in that group — especially among infants who are already medically compromised (e.g., by birth asphyxia, congenital heart disease or hypoglycemia). NEC is almost exclusively seen in newborns. The consensus has long been that intestinal dysbiosis may contribute to NEC risk (among other risks), breastmilk helps, possibly because it can contain probiotics; and probiotics may help, too. Large trials are needed to assess probiotics’ safety and efficacy, particularly in vulnerable subgroups like extreme premies. But existing evidence suggests probiotics reduce mortality and morbidity in this subgroup.
Anyway, Murungi says the PRIMES team doesn’t plan a follow-up “because of the logistical implications.” The incentives are wrong for leading breastfeeding researchers to explore giving sub-Saharan African infants more formula to prevent growth problems from hunger. Trying to solve this problem requires more radical change to the current infant feeding paradigm than scientists invested in it can afford. Maybe Doherty et al did succeed in silencing the opposition, after all.
The current infant feeding paradigm needs to be burned down and rebuilt from the ground up based on science instead of myth in order to prevent common and potentially lethal harm to newborns. But leading scientists can’t say that, embedded as they are within their social and political contexts. It’s not in their self-interests. They don’t offer an empirically well-founded causal story for why widespread growth impairment in sub-Saharan African infants under exclusive breastfeeding promotion is ok, or why feeding babies more is not the solution. They just defend their turf and call it science. When it gets too challenging, they work on other problems that don’t require questioning the entire pseudoscientific edifice on which they’ve built careers.
Tangentially, Flaherman et al didn’t report enough about how they handled missing data due to dropout (by using unspecified linear mixed models) for it to be subject to critical evaluation. Reviewers and journal editors should have required this. The infrastructure is there: PRIMES has an Open Science Foundation page where such code and analyses could be publicly archived, but aren’t. The common problem to check for here is that researchers frequently misuse linear mixed models by failing to include covariates that may predict dropout. This violates the chief assumption that data are Missing At Random (MAR).
Scene 2: Extreme premies fighting for their lives (mortality c. 15%). Top US academic medical centers for newborns. A house of sticks called MILK.
The second little pig wasn’t trying to contest the dominant infant feeding paradigm, not even in an incrementalist way. It happened by accident. In “Neurodevelopmental Outcomes of Extremely Preterm Infants Fed Donor Milk or Preterm Infant Formula: A Randomized Clinical Trial” (aka the MILK trial; JAMA, 2024;331(7):582-591), Colaizy et al looked at whether donor milk versus formula improved extremely preterm infants’ cognitive development and other outcomes. They stopped the trial early, under-powered; they had calculated that they needed 670 infants to detect a 5-point difference in the primary outcome of interest (the BSID cognitive score) with 80% power, but only randomized 483 infants. The early results showed a possible mortality benefit in the formula-fed group — fewer formula-fed babies were dying than donor breastmilk-fed babies.
But that’s not why the researchers stopped enrollment early. They stopped it early because ascendant “breast is best” ideology in top U.S. neonatal medical centers during the study period (Sept. 2012 through March 2019) meant more centers were using donor breastmilk to feed these premies, and there was “declining enrollment and loss of equipoise among participating centers” (p. E4). It is not explained what is meant by loss of equipoise; equipoise is generally defined as scientists’ belief in non-superiority between treatment and control group. But the totality of evidence presented in this article, and the scientific discourse surrounding it, implicate confirmation bias favoring breastmilk over formula. Ideology subverted science.
If it had been run as planned, this trial could have produced important information on a number of open questions. Underpowered as it was, it can only raise the question of whether formula-feeding extreme preterms might increase their chances of survival. That wouldn’t be so surprising…
A range of scientific and clinical reports have long suggested that babies need to gain weight early and often, formula boosts that weight gain, and the most vulnerable babies (i.e., preterm and low birthweight ones like these) may benefit most from these gains. The authors themselves note “Donor milk has been associated with worse growth outcomes than preterm formula” (Colaizy et al, E7). Their trial found “Mortality (death prior to follow-up) was 13% (29/231) in the donor milk group vs 11% (25/233) in the preterm formula group (adjusted risk difference, −1% [95% CI, −4% to 2%])… Weight gain was slower in the donor milk group (22.3 g/kg/d [95% CI, 21.3 to 23.3 g/kg/d]) compared with the preterm formula group (24.6 g/kg/d [95% CI, 23.6 to 25.6 g/kg/d])” (E1).)
In other words, formula supplementation may have resulted in faster weight gain and a mortality advantage for extreme premies. But we don’t know, because the researchers stopped the trial early citing equipoise — in favor of supplementing with donor breastmilk, instead. The article doesn’t say, but in internal discussions and correspondence with their IRB, they probably hung it on increased NEC risk: The infection “occurred in 4.2% of infants (10/239) in the donor milk group vs 9.0% of infants (22/244) in the preterm formula group (adjusted risk difference, −5% [95% CI, −9% to −2%])” (E1). The authors note “Similar reduced risk of necrotizing enterocolitis with the use of donor milk was reported by O’Connor et al and in the Cochrane meta-analysis of formula vs donor milk for pre-term infants (relative risk, 0.53). This finding, which is consistent across multiple studies, suggests that donor human milk, like maternal milk, can reduce the risk of necrotizing enterocolitis compared with formula diets” (E7).
This finding is a red flag. With evidence suggesting breastmilk and probiotics are both protective against NEC, Colaizy et al arguably should have incorporated probiotics into their study design in both arms. They were comparing fortified pasteurized donor breastmilk with formula anyway; both arms got fortifiers in their milk. Why not probiotics to lessen NEC risk, too?
Alternately, they might have run the trial as a 2x2 factorial design with donor breastmilk and formula groups with probiotic and no probiotic subgroups. That would have required more resources — a requirement which may have been prohibitive. At the very least, the authors should have acknowledged the possibility that their subjects should have gotten probiotics to protect them against NEC, and said why they didn’t.
My best guess is the FDA puts out a weird misreading of the literature here, warning “Risk of Invasive Disease in Preterm Infants Given Probiotics Formulated to Contain Live Bacteria or Yeast,” when there are case reports of that — but also a lot of evidence suggesting protective effects. More research is needed to establish net risks and benefits. The Gates Foundation is prioritizing some of it this year, with a 16,000-baby Phase III trial of the breastmilk-derived powdered probiotic B. infantis in breastmilk in five countries.
The bottom line is that, in spite of the increased NEC risk in the formula group, MILK showed a possible mortality benefit for extreme premies associated with formula over donor breastmilk. The causal mechanism could have to do with better nutrition from formula, as reflected in better weight gain. We need to know if that mortality effect really exists. But the incentives are skewed such that, after researchers stopped this trial early, another team in the same field is unlikely to follow up and find out.
In addition to stopping the trial early when the evidence did not support that action, the authors also made methodological mistakes that are pervasive in medical and scientific literature. These include statistical significance test misuse, as evidenced in language including “no effect,” “outcomes were not affected,” and “did not differ,” combined with absence of full interpretation of 95% compatability (aka confidence) interval results; and nullism — arguing that because evidence doesn’t clearly prove a certain, causal link of a particular magnitude between donor milk versus formula supplementation and better cognitive development, that one doesn’t exist. Correcting these errors would change the article’s conclusions.
Ironically, these mistakes caused the authors to miss a chance to promote their own, preferred “breast is best” story. They pre-planned to look for a 5-point difference in the primary outcome of interest (the BSID cognitive score), to consider that clinically significant. Their results show a possible mean language score difference of -1.45 to 5.06 for the donor breastmilk-fed group. In other words, the effect they expected to prove may have actually panned out, at least in part — had they kept running the study as planned.
It’s possible, based on these results, that donor breastmilk conferred a short-term language development advantage, but formula conferred a survival advantage. However, it’s also possible that, if this is the case, the developmental advantage really reflects survivor bias: More donor breastmilk-supplemented extreme premies died, and those who survived had slightly higher toddler language scores on average — because the ones who would have had lower scores were (for reasons of shared underlying causes, such as inadequate nutrition) also the ones who died.
In addition to these common methodological mistakes, Colaizy et al also egregiously mishandled missing data, scoring dead babies as having a fixed cognitive score value in their main analyses. There is no analytical universe in which this is a valid strategy for dealing with this missing data. Only the sensitivity analysis which excludes dead babies from cognitive scoring (Table 3, bottom rows) is valid, and only results from this analysis should have been published.
From a broader perspective, this study design could be criticized for the same reasons Flaherman et al’s and most other science can be: It doesn’t put causality first in asking what’s best for infants, it doesn’t adequately synthesize relevant literature to protect vulnerable populations, and it doesn’t listen to the ordinary people affected by the research (here, the disproportionately poor, black mothers of these extreme premies).
The central causal question here is what optimizes newborn neurodevelopment. Is it nutrition, with formula potentially beating donor breastmilk on vitamins and macronutrients alike? Is it probiotics? Is it something we don’t know or understand well yet about breastmilk?
Both PRIMES and MILK seem to assume, along with the rest of the status quo defenders, that breastmilk is magic. This assumption then tracks with their inadequate synthesis of existing literature, from the small-scale mechanistic, to the large-scale. At the smaller level, probiotics might improve infant health and especially may reduce NEC and other infection and malnutrition risks.
At the larger level, poverty appears to be a root problem in both sub-Saharan African and extreme premie populations — an elephant in the room these scientists seem to ignore. It just takes a sentence to acknowledge “These subjects were born very premature — a hazard of poverty that it would be better for everyone to prevent with adequate social support than to treat with a particular kind of milk.” But that would politicize the issue, and scientists are supposed to pretend to be neutral observers instead of members of society.
So MILK’s methods were flawed to the point of invalidity, the design predictably put some vulnerable infants at increased risk of NEC without acknowledging literature on how to possibly mitigate that risk, and researchers stopped the trial early because ideologues wanted to give all subjects the donor breastmilk supplementation that evidence showed may have put them at a survival disadvantage. The incentives were wrong for breastfeeding experts to observe that, and they didn’t.
Exclusive breastfeeding ideologues did, however, have interest an editorializing on MILK. The trial’s (misinterpreted) null neurodevelopmental findings threatened their preferred story, “breast is best.” So there was pushback in the discourse, with insistence that breast is still best, even if this particular evidence doesn’t show it.
Enter: Wolf.
In the editorial “Delivering on the Promise of Human Milk for Extremely Preterm Infants in the NICU” (JAMA. 2024;331(7):567-569), Mandy Belfort and Maryanne Perrin misrepresent Colaizy et al’s findings and the surrounding medical literature to defend the status quo against this perceived friendly fire. They call for future research following up with the study cohort for later, clinically meaningful outcome differences. Ignoring the evidence of weight gain and survival advantages in the formula group, they emphasize “the MILK study provides no indication of harm to neurodevelopment or other outcomes and adds to already-convincing evidence that donor human milk protects against necrotizing enterocolitis” (p. 586). To be blunt, dead babies are a pretty solid indication of possible harm.
While Belfort and Perrin acknowledge that unmeasured confounding could explain observational studies’ associations between mothers’ own breastmilk intake and better neurodevelopmental outcomes, they then conclude that “Ensuring that donor human milk is available to all extremely preterm infants who need it is an important stop-gap solution. An even bigger priority is to ensure that all extremely preterm infants have access to their mother’s own milk in the NICU” (p. 586). This conclusion is at odds with the evidence. But it was not sui generis.
Belfort and Perrin’s main argument here, that MILK didn’t prove harm from donor breastmilk supplementation, and so breastmilk should be the standard, was echoed in scientific discourse outside the publication record. For instance, Leslie Bienen argued at Sensible Medicine that “ ‘Mechanistic studies on maternal-offspring immune crosstalk show we know little about consequences of not breastfeeding,’ ” Feb. 21, 2024. A MOM (“mother’s own milk”) believer, Bienen writes:
Right now, however, we must acknowledge that we do not understand the long-term effects of not breastfeeding. We know that formula is not killing babies immediately and can be lifesaving. Yet, we understand very little about the long-term implications of foregoing the immune education of infants that happens through breastfeeding.
What’s the posited mechanism of MOM’s magic? Bienen continues:
Other studies show that transfer of IgA, IgM, and IgE occur throughout lactation. IgA induces tolerance to antigens, in addition to providing barrier immunity at the mucosal level, and is completely absent from formula. Little is known about transfer of other Igs such as IgD. The Ig composition of breastmilk changes daily, even hourly, and varies by individual, and is so complex the term ‘immunoglobulinome’ has been coined to describe it. This high variability of breastmilk composition also makes it difficult to meaningfully compare breastmilk to formula. The composition of breastmilk is altered by the mother’s exposure to pathogens in her environment, her sleep, diet, microbiome, by the baby’s growth and caloric needs, and other factors.
Right. Breastmilk isn’t a homogeneous substance. And there’s a lot we don’t know about how its immunological components and other factors can affect kids.
But not all these possible effects are good. To the contrary, some kids may actually learn bad immunological behavior from their mothers’ milk…
Immune Miseducation through Breastfeeding
That’s how a research team in Denmark headed by Hans Bisgaard synthesized an apparently conflicting body of experimental (pro eczema effect) and observational (null or reverse eczema effect) breastfeeding studies to test hypotheses about subgroup effects in a population at high risk for eczema: Babies born to mothers with asthma. Bisgaard was Professor of Pediatrics at the University of Copenhagen, and founder and Head of the Copenhagen Prospective Studies on Asthma in Childhood (CPSAC). (A renowned pioneer of pediatric asthma research, he died unexpectedly at 67 in 2022.)
Bisgaard’s keystone work followed one cohort of mother-infant pairs for an extended period. The cohort was considered high-risk, following children born to mothers with asthma, looking especially for skin and lung-related diseases. In 2010, a team of CPSAC researchers supervised by Bisgaard, Giwercman et al, found breastfeeding protected babies from developing wheezing, but made them more likely to develop eczema. The risk was significant and went up the longer the mothers breastfed.
The researchers guarded against problems like reverse causality (mothers breastfeeding longer thinking it would help their babies with wheezing or eczema) and conditioning on the future by using only information on breastfeeding before disease onset, entering time since breastfeeding as a variable at three-month intervals, and including individuals for whom breastfeeding ended before disease onset. The general method here was imported from efforts to disintangle smoking's causal effects on asthma (Troisi 1995), analyzing concurrent exposure and endpoint, though the implementation was bespoke.
Bisgaard’s team’s results were consistent with their interpretation of the previous body of mixed results, and supported their theory that something in these high-risk mothers’ breastmilk was increasing their babies’ eczema risk in a dose-responsive way. They had suspected the fatty acid content of the milk might matter, and the effect went the way they had suspected (total omega-3 fatty acid content was protective), except where it did not (total omega-6 did not look harmful) — but it was non-significant, possibly due to limited statistical power due to small sample size.
Once they looked at the subgroup of the subgroup of interest (infants who developed eczema while exclusively breastfeeding), they really still needed more data. They held out eczema cases diagnosed before 3 months, which might have biased their results by excluding a particular phenotype. Overall, they worried their results might not generalize beyond their select subgroup (white, high-risk, selected against being born too early or unhealthy).
But their results suggested it was worth thinking more about this ostensible magical substance, so rich in mysterious immune factors and under-studied substances. Were all components of breastmilk benign and desirable, after all? What if some of the pro-inflammatory cytokines turned out to be, well, pro-inflammatory? What if mothers could pass on components of their immune dysregulation or inflammatory disorders through their milk? What if average treatment effects (“breast is best”) were hiding subgroup effects where sometimes breastfeeding adversely impacted pediatric health, depending on the outcome (wheezing versus eczema), the milk's fatty acid or cytokine content, or something else?
Bisgaard and his colleagues continued their investigations. They found the early effects attenuated by age 7 (Jelding & Dannemand 2015): Duration of exclusive breastfeeding did not matter long-term for children's eczema, wheezing/asthma, or allergic rhinitis risks. They found a particular immune factor, interleukin 1 beta (IL-1β) — a cytokine protein — in breastmilk was associated with a greater than 50 percent reduction in eczema risk before age three (Jepsen 2016).
But they had measured 19 immune substances in 223 breastmilk samples to find the result. So they performed a statistical correction for the high probability of false discovery. That method of discovery, however, still leaves open to question whether the finding was due to chance. And they could not say whether the cytokine itself had a protective effect, or just proxied for a healthy maternal immune system that protected babies in some other way.
This was overall consistent with the story that breastmilk is not one things; it varies, and so breastfeeding could affect different subgroups differently. Perhaps mothers with more inflammation or immune dysfunction could pass that on through their milk. That would have enormous practical implications; it might mean that women with such disorders could actually put their children at risk by following current guidance to breastfeed.
There were other clues in this subcorridor of a subcorridor that this might be the case. In 2012, Bisgaard's research group (Carson 2012) replicated a 2004 finding (Linneberg’s) that, in a cohort of high-risk infants (i.e., at least one parent with allergic disease or asthma), maternal alcohol consumption during pregnancy significantly increased a child's risk of developing eczema. This might be consistent with an inflammation-related fetal programming story. Some medical literature asserts that alcohol promotes inflammation, particularly in the brain and intestines (Cannon 2018; Wang 2010)...
So an alcohol-eczema connection could be consistent with the breastfeeding-eczema connection in terms of a possible inflammation and fetal programming story. That story did not go viral. In a typical 2018 review article entertaining the possibility of fetal programming effects of breastfeeding in the context of autoimmunity, Borba et al missed these and any other findings that might have introduced the possibility of harm from some breastmilk to some babies. They did not discuss eczema — an odd omission considering it’s the most common pediatric autoimmune problem.
But to their credit, they do identify another world of breastfeeding risks: Breastfeeding depends on elevated prolactin in nursing mothers, and elevated prolactin is strongly associated with several systemic autoimmune diseases common in reproductive-age women. On that basis, they suggest there is sufficient evidence to recommend that women with systemic lupus erythematosus not breastfeed, and say it’s an open question due to insufficient subgroup evidence whether the association between hyperprolactinemia and other autoimmune disorders (Type 1 diabetes, celiac disease, Addison’s disease, Grave’s disease, Hashimoto’s thyroiditis, multiple sclerosis, uveitis, lymphocytic hypophysitis, reactive arthritis, rejection of heart transplant, rheumatoid arthritis, systemic sclerosis, and Sjögren syndrome) means the same thing. But there is no relevant professional association (e.g., of rheumatologists) that issues guidance of this nature, advising women in those subgroups to consider not breastfeeding for medical reasons.
The story is still unfolding. Perhaps high-risk mothers (e.g., those with asthma, allergies, and autoimmune or metabolic disorders) can minimize fetal exposure to inflammation in pregnancy by avoiding alcohol and ultraprocessed foods, eating unprocessed whole food diets heavy on fruits and vegetables, not smoking, exercising, sleeping, taking fish oil, and/or the other usual suspects. Could they also limit or avoid their babies logging further inflammatory or immune dysfunctional maternal exposures by formula-feeding? Should they?
The consensus position is a resounding no; “breast is best.” But that position assumes myriad benefits of breastfeeding to mothers and children — benefits the evidence does not establish.
Bisgaard’s high-risk mothers were asthmatic — a large subgroup, nearly 10 percent of adult women. But there are many other, smaller subgroups who might be considered high-risk in this context for similar reasons, like women with autoimmune disorders, which are individually rare but common in the aggregate; women with metabolic problems, which are increasingly common in the context of the diabesity epidemic; and women with depression, which also correlates with inflammation for probable bi-directional causal reasons (as discussed here).
Bisgaard’s revelation that some breastmilk seemed to be worse for some babies than formula was the cutting edge of breastfeeding medicine — still searching, leading away from the breastfeeding promotion trial PROBIT’s widely misinterpreted results, and so hyperspecialized as to be largely ignored by researchers working in closely related areas… As often happens in a sprawling, rapidly growing medical literature dominated by researchers with perverse incentives to “publish or perish.”
Overall
In this discourse, Colaizy et al and their (guild-internal) critics alike promote the dominant “breast is best” narrative that is based on myth, not science. There is an alternative narrative in which we admit we don’t know what’s best, aspire to make potentially life and death decisions for vulnerable populations based on science rather than myth, and listen to the people most affected about what’s best for them in the process.
Notably, the African-American subgroup overrepresented in the extreme premies population as in this study is also at heightened risk of serious harm (e.g., kernicterus) from jaundice. African-American mothers themselves disproportionately opt to formula-feed — a choice that confers some protection against these risks, because formula speeds bilirubin excretion. Maybe most disproportionately African-American mothers of extreme premies don’t want to breastfeed them; and maybe they’re right in the sense that their preferences are consistent with what’s medically best for their children.
There was, arguably, no good reason to do this study in the first place. There is no evidence establishing a causal benefit of breastfeeding on cognitive development. That is a myth leftover from a swamp of confounded observational studies and PROBIT, the cluster randomized trial that wrongly legitimated them by failing to consider breastfeeding problems and accidental starvation as possible factors contributing to which infants wound up getting formula, as well as later health and cognitive outcomes. The best available evidence could be interpreted as showing that it is early nutrition, not exclusive breastfeeding, that improves outcomes. This would mean that ensuring adequate early nutrition, not breastmilk intake, should be the first priority of infant feeding guidelines. This would mean putting infant health before breastfeeding.
There is still good reason to compare donor breastmilk and formula plus probiotics in extreme premies in a 2x2 trial — to establish whether formula confers a survival advantage. But it’s not in breastfeeding ideologues’ interests to run a trial that could generate evidence undermining the rationale for promoting supplementing premies with donor breastmilk.
Scene 3: Young infants (one to four month-olds). Top children’s hospitals in Moscow. A house of bricks (normal science serving corporate interests).
The third little pig built a house out of proving something we already knew that would advance his company’s monetary interests. In “Effect of Lactobacillus reuteri NCIMB 30351 drops on symptoms of infantile functional gastrointestinal disorders and gut microbiota in early infants: Results from a randomized, placebo-controlled clinical trial” (Eur J Pediatr, Vol. 183, p. 2311–2324, March 2024), Tyrsin et al looked at a 25-day exposure to a particular strain of a probiotic (L. reuteri) that sometimes occurs in breastmilk and is well-known for promoting infant and child health, especially gut health. As expected, they found it helped babies throw up less, poop more if they were constipated, poop less if they had diarrhea, and cry less.
Specifically, exposed infants
had significant reduction in the numbers of colic (change from baseline - 6.3 (7.34) vs - 3.0 (7.29) in placebo, P < 0.05) and numbers of crying cases and mean duration of crying (decrease from baseline - 144 (70.7) minutes, lower in the diarrhea subgroup than in constipation infants, compared with - 80 (58.9) in placebo, P < 0.0001), as well as regurgitation numbers (decreased by - 4.8 (2.49) with L. reuteri vs - 3 (7.74) with placebo). We also observed increased numbers of evacuations in infants with constipation (L. reuteri 2.2 (2.4) vs 0.9 (1.06) in placebo, P < 0.05). There was a remarkable reduction of evacuations in infants with diarrhea, while not statistically significant.
These findings are consistent with a large body of important findings suggesting probiotics are good medicine. Treatments for colic have come a long way since Matthew Perry’s desperate parents gave their screaming baby phenobarbital — an exposure he would later link with the substance use problems that contributed to his early death.
Still, this study design can be criticized on two grounds: First, it violated the principle of equipoise insofar as the best available evidence suggested all infants with symptoms of GI problems should have had the treatment. There’s a good evidence base for probiotics promoting GI health, and even lessening other infection risks (e.g., respiratory infections in kids with GI problems). It’s not clear what the rationale was for doubting this strain would work, or for denying symptomatic babies a well-established treatment.
Moreover, the evidence establishing probiotics’ efficacy for pediatric GI problems isn’t exclusive to L. reuteri. Rather, there’s a wide array of evidence that other strains (especially Lactobacillus and Bifidobacterium) in particular, and probiotics administered in combinations (larger numbers of strains) in general benefit gut health.
This points to the second design criticism: The trial was designed to produce evidence that a particular product worked, not to figure out what probiotic treatment worked best. Given that combination probiotics might work better (e.g., more diversity when you have to battle a difficult bug = better odds beating it), it’s rational to wonder whether single-strain probiotics are more profitable for companies like the one that stands to benefit from this research.
Lead study author Oleg Yu Tyrsi is CEO of Novo Natum LLC. Second author Dmitry Yu Tyrsin is its Executive Managing Director. Novo Natum distributes “natural medicines for children in Russia and the CIS countries.” They probably tested this probiotic in Moscow, because they imported that strain and want to sell it. Might Novo Natum be able to sell a probiotic with one strain about as well they would one with ten strains that costs more to make?
There’s also a problem with the reported results. The trial was preregistered in the U.S. Clinical Trials Registry, where researchers said they planned to look at several outcomes — but have only gone to publish some of them. The published results deal with GI problems and symptoms thereof (colic, constipation, diarrhea, and gastroesophageal reflux). The unpublished results deal with skin problems (atopic dermatitis/eczema).
Moreover, the registry entry name suggests the scientists conceptualized their trial in one way, but then reconceptualized it in light of their results after running the trial. The registered name was “Randomized Placebo-controlled Study of L. Reuteri NCIMB 30351 in GI Functional Disorders and Food Allergy in Newborns.” This suggests the researchers may have been thinking in terms of food allergies causing both GI and skin problems. This framing does not appear to reflect familiarity with the evidence on either causal contributors to infant GI and skin problems, or the development of food allergies — an area recently pioneered by King’s College London Professor of Paediatric Allergy Gideon Lack, in the randomized trials EAT and LEAP. (These studies suggest lack of early exposure (4-10 months) to common food allergens (peanut, wheat, dairy, soy, egg, and tree nuts) causes problems. So much for exclusive breastfeeding guidelines that commonly recommend against introducing complementary foods before six months.)
Maybe Tyrsin et al will eventually publish the missing preregistered results — unveiling a prepublication theory showing their study design accounted for causality in the context of the food allergy literature and parental reports. Maybe there was a good reason to wait to report the missing outcomes data. But the point of preregistering studies is to require scientists to do what they say they’re going to do, to reduce the opportunities for scientists to only publish positive results (publication bias, aka the file-drawer problem), p-hacking (fraudulently manufacturing those positive results), and other practices that degrade the quality of the scientific literature. However, like most ethical and legal norms violations, the burden is on people with stakes in contesting them to do it.
There is no such wolf here. There is no (other) letter to the editor, editorial, or online criticism of Tyrsin et al to be found, arguably because there are no natural enemies of their enterprise. It looks like a good business in fiscal as well as moral terms, importing quality probiotic strains and showing they work for some medical purposes in your local context. The researchers aren’t standing up to Big Diarrhea when they prove that probiotics help babies with GI problems. They’re just doing one-tiny-crank-of-the-wheel normal science to promote their corporate interests. At least it helps some babies.
So it would be easy to argue this study is the least morally reprehensible of the lot: It didn’t knowingly promote underfeeding babies (Flaherman et al). It didn’t stop early in favor of one treatment after showing a possible mortality benefit for the other treatment (Colaizy et al). It still denied the control group a treatment that I wouldn’t have denied my own child based on the evidence. (I gave my baby combination probiotics early and often, and still do now that he’s a threenager.) And it used finite resources confirming something we already knew (probiotics are generally good for tummy troubles; in particular, L. reuter works for colic).
Those resources could have gone to investigating something we don’t know, instead. Like, how do we help babies in sub-Saharan Africa get enough (water, calories, nutrients) in the first six months so that more of them thrive and fewer of them die? How do we help extreme premies gain weight fast and survive without developing life-threatening infections like NEC? These are questions that matter in bigger scientific and social senses than “Can we prove our product helps kids as expected, so we can sell it?”
But incentives favor doing trials that produce expected results, don’t try anything too risky, don’t threaten the status quo, and thus promote investigators’ careers. So that’s mostly what they do. And, since career scientists with subfield expertise and stakes are in charge of what trials to run, that’s how it stays. Unless…
Scene 4: Newborns with weird (and mostly WEIRD) caretakers. The Internet. A glass house (an idealized trial).
Unless someone makes good on this possibility I keep writing about, of “a platform that helps people conduct surveys and randomized experiments,” bolstering “the ability of citizen scientists to set the research agenda, work towards answering open empirical questions, and create competitive pressure for scientific institutions to cut the administrivia and get with the open science program.” Then, trials like the one I envision might be possible.
In previous research, I proposed an infant feeding “target trial” cluster-randomizing a diverse group of hospitals to transition sooner (treatment) or later (control) to a prelacteal feeding alternative to the exclusive breastfeeding paradigm. According to the Cochrane Handbook chapter on assessing the risk of bias in a non-randomized study, a target trial “is a hypothetical pragmatic randomized trial of the interventions compared in [a systematic review and meta-analysis], conducted on the same participant group and without features putting it at risk of bias” [72]. The Handbook suggests meta-analysis authors start by describing such a target trial. In this context, I was interested in neonatal jaundice and neurodevelopmental outcomes.
Moving up now to infant feeding more broadly, I would look to minimize risk of bias from variations in breastmilk sufficiency (is the mother making enough milk?), immune factors, nutritional, and probiotic content by comparing not breastmilk versus formula, but rather an infant feeding approach promoting exclusive breastfeeding (“breast is best”) versus one prioritizing nutrition (“fed is best”). The trial would be powered to find differences in easily measurable infant and child health and developmental outcomes like infections and delays. It would compare outcomes from both groups with or without probiotic supplementation in a 2x2 factorial design.
Existing evidence including from PROBIT cannot disambiguate between harms associated with accidental starvation when breastfeeding fails, and benefits of The sad reality of breastfeeding science is that it’s impossible to randomly assign women to breastfeed or formula-feed their infants. Common breastfeeding problems like insufficient milk drive formula use and breastfeeding cessation, and women who can exclusively breastfeed their infants systematically differ from women who can’t — so selection effects might explain apparent breastfeeding benefits. That doesn’t mean we can’t run randomized trials on breastfeeding, but it does mean we can’t treat breastfeeding promotion as a proxy for breastfeeding, as PROBIT investigators did. That introduced risk of bias.
This trial design, by contrast, would permit assessment of whether an infant feeding approach that promotes nutrition or one that promotes exclusive breastfeeding generates better outcomes. Promoting nutrition would include encouraging prelacteal, supplementary, and complementary feeding, as opposed to telling mothers to feed their babies only breastmilk for the first six months. Both groups would get the same standard advice on initiating breastfeeding and administering vitamin D supplementation, which are universally recommended.
Target trials are supposed to dream big, so let’s say we want data on infant hunger cues, GI problems, colic, infections, and developmental outcomes. Ideally, we would also want to record mothers’ perceptions of how breastfeeding was going, and measure how much breastmilk babies got — a difficult target, since breastfeeding proponents often argue pumps don’t get milk out as efficiently as babies, and bottle-feeding breastmilk isn’t the same as breastfeeding. We would also like to know about the breastmilk’s changing composition over time in terms of macro and micronutrients, milk being often nutritionally poorer the longer a mother has been lactating.
Practically, however, trial design is also supposed to be pessimistic. This data collection would be prohibitively difficult and resource-intensive; even logging formula amounts consumed and diapers filled might prove too much for a lot of new parents. But if we just got data on different feeding approaches and health outcomes, that could generate evidence on which to remake infant feeding guidelines. In the real world, we may have to trade some risks of bias for some chance of generating evidence.
Another risk of bias in studies comparing formula versus breastmilk supplementation (e.g., Colaizy et al) or formula supplementation versus exclusive breastfeeding promotion (e.g., Flaherman et al) comes from potentially variable probiotic content in breastmilk including donor pasteurized breastmilk. Although pasteurization should kill good bacteria along with bad, it could fail to; or good bacteria could tag along in non-pasteurized human milk-based fortifiers commonly added to donor breastmilk. And, although breastmilk can have some good bacteria, the type(s) and amounts vary.
So an idealized infant feeding trial would assess background probiotics exposure, measuring probiotics in breastmilk to gage variation, and systematically vary probiotics exposure — including at least one probiotic exposure arm. But, realistically, measuring breastmilk probiotics content might be prohibitively resource-intensive.
A more practical and ethical alternative might be to equalize baseline exposure by simply giving all subjects some probiotics, and varying what type (single or combination strain). This approach would, however, likely compromise the trial’s ability to generate evidence on associated morbidity/mortality rates that seems needed to change some minds (e.g., the FDA’s) on the safety of probiotics in infants. It would also tend to make detectable differences much smaller and thus less publishable. Science is structured to incentivize detecting big differences and publishing on them.
So even researchers who believe the best available evidence already favors giving all infants probiotics are likely to structure research denying some subjects that treatment in order to prove its safety and efficacy. There are longstanding debates on the practical and ethical dimensions of this type of quandary, with proponents of randomized trials invoking the need to further diminish uncertainty about what we really know, and proponents of equalizing access to treatments that probably work emphasizing principles of justice. The answer here hinges on how good the probiotics evidence really is — a matter of interpretation about which rational, well-informed people can disagree.
In addition to adding probiotics to the trial design, incorporating combination strain treatment and no probiotics control arms in one iteration (a better experiment if you’re not convinced of probiotics’ safety and efficacy) or combination versus single strain arms in another (a more ethical experiment if you are), I would change the venue from hospitals to infant caretakers at home. Most moms and babies don’t stay in the hospital long after birth. So if we want to know how different infant feeding approaches in the first six months affect infant growth, health, and development, we need to try them with people at home.
Yes, it would be a select subgroup of caretakers interested in putting their babies in this kind of a trial. Mostly moms; mostly WEIRD (white, educated, industrialized, rich, democratic country-context) folks. Randomized trials usually involve select populations, and generalizability to other subgroups (e.g., sub-Saharan African contexts, or extreme premies) would have to be assessed in light of novel evidence in such groups. There are enough open empirical questions about infant feeding that a large-scale trial without country context restrictions could generate needed information to improve infant feeding recommendations currently lacking in evidentiary basis.
The Human Element
I would be remiss if I didn’t (again) point out the context of my own interest in this topic, which originated in accidentally starving my newborn for over a month under the auspices of exclusive breastfeeding. Healthcare professionals encouraged me to keep it up, when it clearly didn’t work. Leading experts told me clearly crazy things, like that it’s fine for exclusively breastfed babies to not poop for extended periods of time. (Sufficient food in creates waste out; breastmilk does not magically disappear without waste, no matter how much you may believe in its other, ostensibly magical powers.)
And so I began to realize, like other moms with similar experiences, that the medical and scientific recommendations on infant feeding were entirely out of touch with the reality of putting my child’s health first by ensuring his nutrition. I keep returning to the infant feeding literature now to see if it’s gotten any closer to preventing other infants suffering the common harm that my son did. While the mainstream conversation is set to keep making strides in this direction, the scientific conversation lags. Perverse incentives keep it stagnant, still waiting for a massive course correction that is far overdue.