Part Two / The First Year

Breast Is Best?

What the strongest evidence does and does not show about breastfeeding.

Chapter 04 · 9 min full read

In this note

The one thing to remember

Breastfeeding has some real, modest, short-term benefits for the baby and a likely meaningful long-term benefit for the mother. The dramatic long-term claims — IQ, obesity, allergies, lifelong health — do not survive the studies designed to test them. The gap exists because in the US, the women who breastfeed are systematically different from those who don't.

Cutting through the noise#

This is the chapter where the book's method matters most, so it's worth understanding the problem before the answers.

Nobody breastfeeds at random. In the US and most wealthy countries, breastfeeding is more common among women with more education, higher income, better maternity leave, and more support at home. Every one of those things independently predicts better outcomes for children.

So when a study finds that breastfed kids do better in school, you have two candidate explanations, and the data alone can't separate them:

  1. Breast milk did something.
  2. The kind of family that manages to breastfeed was already going to produce that outcome.

Researchers try to fix this by "adjusting for" differences — controlling for income, education, race. But they can only adjust for what they measured, and they almost never measure the mother's IQ, her temperament, or how much she wanted to breastfeed in the first place. Mothers with higher IQs are more likely to breastfeed even among women with identical education levels — and they also have children with higher IQs. That residual difference is invisible to the adjustment, and it goes straight into the "breastfeeding benefit."

One study makes this vivid. The authors analyzed the same children three ways:

AnalysisResult
Simple comparison, few controlsLarge IQ difference favoring breastfed kids
Same data, now adjusting for the mother's IQMuch smaller, but still present
Same data, comparing siblings — one breastfed, one not, same motherNo significant effect

The sibling comparison is the powerful one, because it holds constant everything about the family that you can't measure. Same parents, same house, same income, same genes-ish. And the effect disappears.

How to read the evidence in this chapter#

SourceWhat it isWeight
PROBITA randomized trial in Belarus, 17,000 mother-infant pairs. Half the mothers got breastfeeding support and encouragement; half didn'tHighest. The only large randomized evidence we have
Sibling studiesCompare kids in the same family treated differentlyHigh. Removes family-level confounding
Large observational studies with rich controlsE.g. the Danish ear-infection study of 70,000 womenModerate. Useful for small effects a single trial can't detect
Case-control studiesUsed for rare outcomes like SIDS and childhood cancerLow. Comparison groups are recruited differently and parents recall the past

One caveat on PROBIT that cuts the other way: it was an encouragement trial, not a forced one. At three months, 43% of the encouraged group were exclusively breastfeeding versus 6% of the others — a real difference, but not 100% vs. 0%. So measured effects are diluted, and a small true benefit might not show up as statistically significant. This is why it's reasonable to also look at very good observational data rather than treating one trial as the final word.

The 60-second recap#

  • Raw comparisons of breastfed and formula-fed children measure the family, not the milk.
  • Randomized and sibling evidence consistently shrinks the claimed benefits.
  • What survives: fewer gastrointestinal infections, less eczema, probably fewer ear infections. All short-term, all modest.
  • What doesn't survive: IQ, obesity, allergies, asthma, height, blood pressure, cavities.
  • The maternal breast-cancer reduction is the most robust benefit in the whole chapter — and it's a benefit to you, not the baby.
  • Costs matter and are usually left out of the conversation entirely: pain, time, pumping, sleep, work, mental health.
  • Guilt has no evidence base and a real cost.

What to actually do#

OPTION - ModerateBreastfeed if the real benefits plus your own experience make it worth it
Why this call

The benefits are real but small. That makes this a preference decision, not an obligation

OPTION - HighFormula feed if you prefer, where safe formula and clean water are reliable
Why this call

No compelling evidence of major long-term harm. (In settings where formula must be mixed with contaminated water, this calculus is completely different — that's not the population the chapter addresses)

RECOMMENDED - ModerateIf you want to breastfeed, get practical support — leave, help, a lactation consultant
Why this call

Support often determines whether it's feasible at all. See Chapter 5

RECOMMENDED - ModerateFactor in the maternal breast-cancer benefit
Why this call

It's the best-supported item on the list and it's routinely left out of a decision framed entirely around the baby

SKIP - HighSkip the guilt, and skip claims that feeding method determines intelligence, attachment, or success
Why this call

One woman was told by her doctor that stopping breastfeeding cost her child three IQ points. There is no evidence for that

SKIP - HighDon't treat an unadjusted correlation as proof of causation
Why this call

This is the single most common error in breastfeeding coverage

The claims, checked#

The commonly circulated list of benefits is enormous — it includes "better friendships." Here's what holds up.

ClaimVerdictThe evidence
Fewer GI infections (diarrhea)Likely realPROBIT: 13% of the non-encouraged group had at least one episode vs. 9% of the encouraged group
Less eczema and rashLikely real, short-termPROBIT: 6% vs. 3%. Halved in relative terms, but on a small base
Fewer ear infectionsProbably, smallNot found in PROBIT, but a careful Danish study of 70,000 women found 7% → 5% through six months. A similar UK study found nothing. On balance, plausible
Less necrotizing enterocolitis in very premature babiesRealShown in randomized trials. Important, but NEC is vanishingly rare in full-term babies
Fewer colds, coughs, respiratory infectionsNot supportedPROBIT found near-identical rates. Observational studies here are small and fragile
Lower SIDS riskUncertainCase-control studies find an association, but it shrinks or vanishes once you account for smoking, prematurity, and other risk factors. Studies with better-matched control groups find no effect
Higher IQNot supportedSibling studies find nothing. PROBIT found no overall IQ or school-performance effect; a verbal-IQ signal appeared but the assessors knew which group children were in
Lower obesityNot supportedPROBIT found no effect at age 7 or ~11. Sibling studies find effects between families but not within them — which points at the family, not the milk
Fewer allergies, asthma, cavities; better height, blood pressureNot supportedPROBIT followed children to age 7 and found no differences
Lower maternal breast cancerLikely realConsistent across many studies and countries, roughly 20–30% reduction, with a plausible biological mechanism
Weight loss for momReal but trivialAbout 1.4 lb more at six months in one large study, and that's likely an overestimate
"Free birth control"NoYou are less likely to conceive, but this is not reliable contraception. Use actual contraception
Lower postpartum depressionNot supportedCausality runs both ways: depressed mothers are more likely to stop breastfeeding, which makes breastfeeding look protective
Better friendshipsNo evidence existsNot refuted — simply never studied. It's on the list anyway, which tells you something about the list

On the "not supported" verdicts: this doesn't mean these things are proven false. It means nothing in the good data suggests them. As Oster puts it, you could equally plausibly link breastfeeding to being a fast runner or good at the violin. You can believe it on faith; you just can't call it evidence.

Numbers in context#

  • GI infection: 13% → 9%. A real reduction, and a meaningful one as a share of cases. Also: even in the less-breastfed group, 87% of babies had no episode.
  • Eczema: 6% → 3%. Halved — but the base rate was already low, and these are minor conditions.
  • Breast cancer: 20–30% reduction. This one is large in absolute terms, because breast cancer is common — roughly 1 in 8 women. A 20–30% relative reduction on a common outcome is worth more than a 50% reduction on a rare one.
  • Weight loss: 1.4 lb at six months. Yes, nursing burns calories — but nursing mothers also eat more. Oster reports a daily 10:30 a.m. egg and cheese bagel policy, which more or less guarantees you replace what you burn.

The decision, done properly#

  1. Start from the short list of credible benefits, not the long list of claimed ones.
  2. List the real costs honestly: pain, time, pumping logistics, sleep, medication questions, work constraints, mental health. These get systematically under-weighted because the conversation is framed entirely around benefits.
  3. List the real personal benefits: closeness, enjoyment, convenience, cost, cultural meaning. These are legitimate inputs, not rationalizations.
  4. Decide what's sustainable — exclusive breastfeeding, combination feeding, exclusive pumping, or formula.
  5. Set a point to revisit. This is a decision, not a vow.

Why cognitive dissonance makes this conversation so hostile: once you've made a choice, it's tempting to decide it was the perfect one. If you didn't breastfeed, you don't want to admit there are any benefits. If you spent two years nursing, you need it to have mattered enormously. Both positions are defenses, and both are why the internet on this subject is unbearable. Your choice can be right for you without being right for anyone else, because your constraints aren't theirs.

A useful older voice: Oster looked up what Dr. Spock said in the 1980s edition her mother used. He suggested mothers try it and see if they liked it, mentioned possible infection protection briefly, and then said the most convincing evidence comes from mothers who describe the satisfaction and closeness of it. That's a defensible position, and it's roughly where the data lands.

Where the evidence runs out#

  • The true effect on SIDS, asthma, allergies, and several rare outcomes — the studies just aren't good enough to say.
  • Whether very small long-term effects would survive if we could perfectly account for family differences. Probably unknowable.
  • Any evidence-defined "optimal duration" that applies to everyone. There isn't one.

Our family's notes#

  • Benefits that actually matter to us:
  • Costs and constraints that actually matter:
  • Our preferred feeding plan:
  • Our acceptable fallback, decided in advance:
  • What would make us revisit this:

Find a note

Guidance & evidence

The notes separate an action's strength from confidence in its evidence. These are the study guide's labels, not an independent clinical assessment.

MUST
Safety-critical or medically urgent.
Evidence leans clearly one way.
OPTION
Your circumstances and preferences decide.
TRY
Low risk; worth a limited experiment.
SKIP
Little value for the effort.
AVOID
Evidence of harm or a risk not worth taking.

High confidence: strong and consistent. Moderate: credible, with real limitations. Low: small, mixed, or indirect evidence.