The Breastfeeding Benefits That Survive a Randomized Trial

The largest randomized trial of infant feeding ever conducted found no effect on childhood obesity. None at six years, none at eleven, none at sixteen. This is not a fringe finding and not an argument against nursing — it comes from the most rigorous evidence we have about breastfeeding benefits, and what it says is more interesting than the pamphlets are. Some effects are large and mechanistically clear. Some are real but smaller than advertised. And a few of the most-repeated ones look like artifacts of who chooses to breastfeed in the first place.

The rest of this is me earning that paragraph.

What the trial actually did

PROBIT — the Promotion of Breastfeeding Intervention Trial — ran in Belarus starting in 1996. Michael Kramer and colleagues randomized 31 maternity hospitals and their affiliated pediatric clinics, not individual mothers, to either the WHO/UNICEF Baby-Friendly Hospital training program or the standard care already in place. 17,046 mother–infant pairs enrolled.

You cannot ethically assign a woman to formula-feed. So the trial randomized the encouragement instead, and the encouragement worked: at three months, 43.3% of the intervention group were breastfeeding exclusively, against 6.4% of controls. That gap is the whole experiment. Everything measured afterward is the downstream consequence of a large, randomly assigned difference in how these infants ate.¹

At twelve months (Kramer et al., 2001, JAMA), the intervention group had meaningfully less gastrointestinal infection — 9.1% versus 13.2% — and less atopic eczema, 3.3% versus 6.3%. Respiratory infection did not budge. At 6.5 years, a separate analysis in the BMJ (Kramer et al., 2007) found no reduction in asthma or allergic sensitization. At 11.5 years (Martin et al., 2013, JAMA Pediatrics) there was no difference in body mass index, fat mass, or waist circumference.

So: strong protection against gut infection in infancy. Nothing detectable, in this trial, for the two outcomes most often used to sell nursing to Western parents.

Does breastfeeding make babies smarter?

Probably a little, and the size of the effect is genuinely contested. PROBIT's cognitive follow-up at 6.5 years (Kramer et al., 2008, Archives of General Psychiatry) found the intervention children scored about 7.5 points higher on verbal IQ and roughly 6 points higher on full-scale IQ, with teacher ratings pointing the same direction. That is a very large number — larger than most researchers in the field consider plausible — and it comes with a wrinkle: the testing pediatricians worked within the same clinic clusters they were assessing, which is not the same thing as blinding.

The same cohort was tested again at sixteen (Yang et al., 2018, PLOS Medicine), using different instruments, and the advantage had largely washed out. My reading is that early breastfeeding is associated with a small cognitive edge, that the 7.5-point figure is an overestimate, and that anyone quoting a precise IQ gain to you is more confident than the data.

The parts that aren't in dispute

The immune story is where the biology is cleanest, because you can trace it step by step.

How an antibody gets from your gut into your baby's

A mother swallows a pathogen — say, a rotavirus circulating through her household. Immune cells in the Peyer's patches of her small intestine sample it. B cells there are activated, mature into antibody-secreting plasmablasts, and switch on a surface receptor called CCR10. Those cells then leave the gut and enter circulation.

Meanwhile, under the hormonal conditions of lactation, the epithelium of the mammary gland is releasing a chemical signal called CCL28 — the ligand CCR10 binds. The plasmablasts follow that gradient and take up residence in the breast, where they begin producing dimeric IgA against the pathogen the mother's gut just met. That IgA is ferried across the mammary epithelium by the polymeric immunoglobulin receptor, which clips off a fragment called secretory component and leaves it attached.

That fragment is the point. It makes the antibody resistant to the infant's stomach acid and digestive proteases, so the molecule arrives in the small intestine intact. It is not absorbed into the baby's bloodstream. It coats the gut lining, binds pathogens, and is excreted. Colostrum can carry more than 5 grams of it per liter; mature milk runs closer to 0.5 to 1 gram. An infant's own secretory IgA production is minimal for the first several months, which is precisely the window this covers.

The sugars that aren't for the baby

Human milk contains 5 to 15 grams per liter of complex sugars — human milk oligosaccharides, roughly 150 identified structures — that the infant has no enzyme to break down. They are the third most abundant solid component of milk, and they pass through the small intestine untouched. They are food for a bacterium: Bifidobacterium longum subsp. infantis, which carries the gene cluster to metabolize them and outcompetes less welcome colonizers when they are on the menu.

Some of these sugars also work as decoys, presenting the same surface structures that pathogens use as docking sites, so the pathogen binds the sugar and washes out. Cow's-milk-based formula contains almost none of this; a handful of synthesized HMOs have been added to some products, against a native repertoire of a hundred and fifty.

The mother's side of the ledger

The most robust maternal finding comes from the Collaborative Group on Hormonal Factors in Breast Cancer (2002, The Lancet), a pooled reanalysis of 47 studies across 30 countries covering 50,302 women with breast cancer and 96,973 controls. Relative risk of breast cancer fell 4.3% for every 12 months of lifetime breastfeeding, on top of a 7.0% reduction per birth. It is observational, but it is dose-dependent, consistent across wildly different populations, and biologically coherent — fewer ovulatory cycles, and terminal differentiation of breast tissue.

Why the observational numbers are so much bigger

In high-income countries, women who breastfeed longer are, on average, wealthier, more educated, less likely to smoke, and more likely to have paid leave. Every one of those independently predicts child health. Cohort studies adjust for them and still can't fully separate them.

Colen and Ramey (2014, Social Science & Medicine) tried a sharper instrument: comparing siblings within the same family who were fed differently, using 1,773 discordant sibling pairs from the NLSY. Across eleven outcomes — BMI, obesity, hyperactivity, reading and math scores, and others — the advantages that appeared between families mostly vanished within them.

This is also why global mortality figures are so much larger than trial effects. Where water isn't safe and formula is diluted to stretch it, the hazard is the water and the dilution. That is a real and enormous benefit. It is not the same benefit being marketed to a reader with a filtered tap.

A note on coffee, since this is that kind of magazine

Caffeine crosses into milk readily, peaking about an hour after the cup, at a milk-to-plasma ratio around 0.5 to 0.8. The relative infant dose at moderate intake lands near 6 to 10% of a weight-adjusted maternal dose. What matters more is clearance: a term newborn's CYP1A2 is immature, and caffeine half-life runs somewhere between 65 and 100 hours, against roughly 5 in an adult. By three to five months it has fallen close to adult range.²

Whether this actually wrecks infant sleep is less settled than the warnings imply. Santos et al. (2012), Pediatrics, followed 885 three-month-olds in the Pelotas birth cohort, where heavy maternal caffeine intake was near-universal, and found no consistent association with frequent night waking. One cohort, one population. But the confident version of this advice is running ahead of its evidence.

An honest rule of thumb

If you are lying awake wondering whether tomorrow's bottle of formula undoes something: it doesn't. The effects that hold up are concentrated in the first months, they are mostly about infection, and they behave like a dose, not a switch. Partial nursing gets you partial protection, which is the ordinary way biology works.

What this looks like at my house

My son is nine months old. I drink my second coffee at the start of a feed rather than after it, on the logic that milk caffeine peaks around sixty minutes later, which lands it in the gap between feeds instead of in his mouth. I have no trial for that. The transfer pharmacokinetics are solid; the sleep data are not.

It costs nothing, so I do it. That is the entire protocol: one cup, taken at the wrong end of a feed on purpose, and no scale in the room.