What does breastfeeding do for baby and mother, and what if it does not work?

For babies. The best test is PROBIT, a trial in Belarus that randomized 31 maternity hospitals to a breastfeeding-support program or usual care. Babies at the support hospitals were breastfed for longer, had fewer gut infections (9.1% versus 13.2%) and less eczema (3.3% versus 6.3%), but no fewer chest infections¹¹. A major Lancet review also found protection against infections and probable reductions in overweight and diabetes, with no link to asthma, blood pressure or cholesterol⁶. An umbrella review of 192 meta-analyses rated the evidence for lower sudden infant death risk as convincing⁷.

For mothers. Breastfeeding is linked to lower breast cancer risk, longer gaps between pregnancies and possibly lower ovarian cancer and type 2 diabetes risk⁶. A later umbrella review rated the reductions in ovarian cancer and high blood pressure as convincing⁷.

How long. The American Academy of Pediatrics recommends exclusive breastfeeding for about six months, then continuing with solid foods for two years or beyond if mother and child both wish².

The caveat. Who breastfeeds is not random; in the US it varies by race and socioeconomic status, so some apparent benefits reflect family circumstances⁹. The IQ gap is 3.4 points overall but 2.6 in studies that controlled for maternal IQ⁸, and in sibling comparisons within families most long-term differences vanished⁹. A verbal advantage seen in PROBIT at school age was much smaller at 16¹². Only four of the 192 meta-analyses reached moderate GRADE quality⁷.

The realities. In a survey of 1,177 mothers, about 60% of those who stopped did so earlier than they wanted, mostly over lactation difficulties, worries about the baby's weight, illness or medicines, and the effort of pumping¹⁰. Support from professionals or trained peers probably reduces the number who stop¹.

Milk-supply aids. Evidence for herbal and prescription galactagogues is low to very low in certainty³. Fenugreek caused adverse reactions in many users in one survey and can interact with warfarin⁴. Domperidone is not approved in the US and carries arrhythmia warnings at higher doses⁵.

If it does not work. Combination or formula feeding is common and regulated; see choosing an infant formula.

PROBIT: the largest randomized test

PROBIT cluster-randomized 31 Belarusian maternity hospitals and clinics (17,046 mother-infant pairs, 96.7% followed for a year) to a Baby-Friendly-style program or usual practice¹¹. Exclusive breastfeeding at three months was 43.3% versus 6.4%, and any breastfeeding at 12 months 19.7% versus 11.4%. Gut infection fell (9.1% versus 13.2%; adjusted OR 0.60, 95% CI 0.40–0.91), as did eczema (3.3% versus 6.3%; OR 0.54, 95% CI 0.31–0.95), but respiratory infection did not (39.2% versus 39.4%; OR 0.87, 95% CI 0.59–1.28). Caveats: the trial randomized support, not feeding; mothers already intended to breastfeed; follow-up was one year in one country.

Cognition: a shrinking effect

Horta's meta-analysis of 17 studies found breastfed people scored 3.44 IQ points higher (95% CI 2.30–4.58), falling to 2.62 (95% CI 1.25–3.98) in studies controlling for maternal IQ⁸. PROBIT's 16-year follow-up (13,557 participants, 79.5% of those randomized) found no significant difference in the prespecified intention-to-treat comparison. With extra adjustment, the intervention group scored 1.4 points higher in verbal function (95% CI 0.3–2.5) and 1.2 in memory (95% CI 0.01–2.4); an instrumental-variable analysis suggested 3.5 points (95% CI 0.9–6.1) for three or more months of exclusive breastfeeding. The authors saw no benefit for overall neurocognitive function and a verbal effect much smaller than at school age¹².

Confounding and sibling comparisons

Colen and Ramey used 25 years of National Longitudinal Survey of Youth data. In standard models, breastfed children aged 4 to 14 did significantly better on 10 of 11 outcomes. Restricted to siblings with within-family fixed effects, estimates for all but one outcome dropped dramatically and lost significance, which the authors read as selection into breastfeeding by race and socioeconomic status⁹. Sibling designs have limits too: they cannot prove breastfeeding does nothing, and parents may feed siblings differently for reasons linked to health.

How strong is the overall picture?

The umbrella review (192 meta-analyses from 62 articles) graded convincing evidence for lower sudden infant death (OR 0.63, 95% CI 0.54–0.73) and allergic rhinitis (OR 0.61), and for lower ovarian cancer (OR 0.70–0.78) and hypertension (OR 0.89–0.93) in mothers. Overweight and type 2 diabetes in adulthood were rated highly suggestive and suggestive. Yet only 4 of 192 analyses reached moderate GRADE quality, and most of this evidence is observational⁷. The Lancet series estimated that near-universal breastfeeding could prevent 823,000 child deaths and 20,000 breast cancer deaths a year, and noted more tooth decay with longer breastfeeding⁶.

When breastfeeding is hard

Odom's survey followed 1,177 mothers monthly through the first year. About 60% of those who stopped did so earlier than they wanted, associated with lactation difficulties, concerns about infant nutrition and weight, illness or need for medicine, and pumping effort¹⁰. The Cochrane review of support (116 trials, over 98,000 mother-infant pairs) found that "breastfeeding only" support probably reduced stopping any breastfeeding at six months (RR 0.93, 95% CI 0.89–0.97) and stopping exclusive breastfeeding at six months (RR 0.90, 95% CI 0.88–0.93), with moderate-certainty evidence. Support from professionals or lay peers looked similarly useful, and four to eight contacts may be best. Blinding was impossible and outcomes were self-reported¹.

Galactagogues: weak and not risk-free

Foong's Cochrane review of 41 trials (3,005 mothers) rated the evidence low to very low certainty. Three small trials of metoclopramide, domperidone and sulpiride (151 participants) suggested higher milk volume (mean difference 63.8 mL, 95% CI 25.9–101.7). Botanical trials, including fennel, fenugreek and ginger, were too inconsistent to pool (I² = 99%), and almost no trial reported whether mothers went on to breastfeed for longer³. Reviews summarized in the LactMed database disagree on fenugreek, one finding a mild effect and another none; among 85 mothers who used it, 45% reported an adverse reaction⁴. For domperidone, LactMed notes that doses above 30 mg a day may raise the risk of arrhythmias and sudden cardiac death, and that Canadian database cohorts found increased risk, though confounding makes those results questionable⁵.

Bottom line (confidence 3 of 5)

Protection against infections, sudden infant death and maternal cancers is consistent, but mostly observational. Cognitive and weight benefits are smaller and entangled with family circumstances. Support helps; milk-supply aids are unproven. Feeding is your decision with your midwife or pediatrician, and a well-supported mother and well-fed baby matter most.