Giving iron before anemia shows no measured cognitive benefit

The stores that run out around six months

A term newborn carries most of the iron it will need for its first months before it ever eats. About 80% of a baby's total-body iron is laid down in the third trimester, a reserve built before birth rather than fed after it.¹ Breast milk does not replenish that reserve once it runs low. Human milk carries about 0.35 mg of iron per liter. An exclusively breastfed infant drinking roughly 0.78 L a day takes in about 0.27 mg of iron daily, almost exactly the adequate intake set for the first six months of life.¹ That match is not generous design. It reflects a limit: in early infancy, fetal stores are doing the work, not milk.

Around four to six months, the reserve is spent and the math changes. The recommended intake does not creep upward. It jumps, from 0.27 mg a day to 11 mg a day for the second half of the first year.¹ Part of that jump is an artifact of how the two numbers are calculated, not a sudden physiologic cliff. The report that sets the figure says so plainly.¹ Still, milk at 0.35 mg per liter cannot supply eleven milligrams a day no matter how much of it a baby drinks. Meeting even a third of that from food takes real volume: a jar of strained meat carries roughly half a milligram to just over one, and a tablespoon of iron-fortified infant cereal carries one to two.¹ That is why fortified cereal, meat, and, for some infants, a supplement carry so much of the load once complementary feeding starts. It is also the six-month fork: the free reserve empties, and how a family fills the gap turns from formality into choice.

Deficiency and anemia are not the same shortage

Iron deficiency, anemia, and iron-deficiency anemia are three overlapping but distinct conditions, and the difference changes what a test result means. Iron deficiency is depleted iron stores, detectable on a ferritin test before it touches red blood cells. Anemia is low hemoglobin, which iron deficiency can cause but so can several other conditions. Iron-deficiency anemia, or IDA, is the overlap: anemia specifically caused by too little iron. In the most recent nationally representative US data, iron deficiency affects about 13.5% of one- and two-year-olds, roughly one in seven. Anemia affects about 5.4%. IDA, the overlap of the two, affects about 2.7%, roughly one in 37.² About half the anemic toddlers in that data were iron deficient. The other half were anemic for some unrelated reason.² A hemoglobin test alone, without a separate iron-status test, will miss the more common condition and can misattribute the anemia it does catch.

The gap also widens with age before complementary feeding and any supplementation catch up. A Norwegian cohort of 284 term infants, cited in the American Academy of Pediatrics' clinical report, found iron deficiency in 4% of six-month-olds and 12% of twelve-month-olds.¹ That tripling within half a year tracks exactly the stretch when stores run out and food has to compensate for the first time. None of this is contested. What a shortage this common does to a baby's brain, and whether treating or preventing it changes that, is where the evidence gets harder to read.

Treatment works. Prevention hasn't, yet.

Correcting iron-deficiency anemia once it exists reliably reverses the delay it caused, shown in a randomized trial that diagnosed the anemia by blood test first.³ Preventing iron deficiency before it starts has not, in the randomized trial built to match a healthy, breastfed infant, been shown to change development at all.⁴ Both results come from randomized, placebo-controlled trials. Neither is a guess. They are not the same question.

The reason this question gets asked at all is a large, persistent observational association. In a Costa Rican cohort study, infants followed over years rather than assigned to anything, infants with moderately severe iron-deficiency anemia scored lower than iron-sufficient peers on multiple developmental measures at age five. Their iron levels had normalized well before the testing.⁵ The same cohort, reassessed at 19 years, still showed an 8-to-9-point cognitive gap in middle-income families, and a gap that widened from about 10 points in infancy to 25 points by adulthood in lower-income families.⁶ A cohort study like this cannot rule out that families differing in infancy iron status also differ in ways that shape a child's trajectory over two decades: nutrition beyond iron, schooling, chronic stress, home environment. The widening, income-linked gap is itself evidence for that alternative explanation: if iron deficiency in infancy acted alone as a fixed biological injury, the gap should stay constant across income groups, not grow where other disadvantages compound it.⁶ The association is real and large. Whether it reflects iron alone, or iron as one marker among correlated disadvantages, a cohort design cannot settle.

Randomized trials answer a narrower question than cohort studies do, but more cleanly. Assign iron or placebo by coin flip, and a later difference cannot be explained by which families were already different going in. The clearest positive result of this kind comes from a trial that enrolled only infants already diagnosed anemic. Idjradinata and Pollitt sorted 126 Indonesian infants, aged 12 to 18 months, by blood test into anemic, non-anemic-but-deficient, and iron-sufficient groups, then randomized the anemic group to ferrous sulfate or placebo for four months.³ Before treatment, the anemic infants scored significantly lower on the Bayley scales than the other two groups. After four months, the iron-treated anemic infants had caught up entirely to the iron-sufficient infants; the placebo-treated ones had not.³ That diagnosis came from blood work, not from a parent noticing fussiness or pallor, which is what turns "anemic" from a worry into a condition a pediatrician can treat.

A Cochrane review of eight such treatment trials in already-anemic children under three found the same split by duration. Six trials (225 children), followed 30 days or less, found no significant benefit, on evidence graded low quality. Two trials followed longer, and results diverged: one found a large gain, the other much less, and the reviewers call the longer-term picture "unclear."⁷ Treating diagnosed anemia works, but not identically at every duration.

Prevention is a different experiment, and it has run at larger scale. In Chile, 1,657 healthy full-term infants, not selected for anemia, were randomized to iron-fortified formula or unsupplemented, cow-milk-based feeding.⁸ The unsupplemented infants processed information more slowly, showed less positive affect and social engagement, were harder to console, and crawled somewhat later, alongside far more anemia by 12 months (numbers below).⁸ That is a real, positive result, but from a population with unusually high baseline risk: nearly a quarter of unsupplemented infants became anemic, well above what a well-resourced, breastfed US household would expect.

The largest test of prevention pools most of this literature together: a 2013 meta-analysis of 35 randomized trials and 42,306 children aged 4 to 23 months, most not selected for anemia at enrollment.⁹ Supplementation cut the relative risk of anemia to 0.61, iron deficiency to 0.30, and IDA to 0.14, every confidence interval well clear of no effect.⁹ But pooled across six trials each, mental and psychomotor scores told a different story. The mean difference was under two points either way, 1.65 and 1.05, and both confidence intervals crossed zero. They are too small and too uncertain to call real.⁹ Supplementation also brought more vomiting, more fever, and a small, temporary dip in growth. The authors' own conclusion: iron supplements are established for blood counts. Development, growth, and side effects remain genuinely open questions.⁹

The trial closest to this question's real infant, healthy, non-anemic, breastfed, in a high-income country, used a purpose-built design and reached the same conclusion. In Poland and Sweden, 221 such infants were randomized to daily low-dose iron or placebo from 4 to 9 months, then tested with Bayley-III at 12, 24, and 36 months. There was no significant difference in motor, cognitive, or language scores at any age.⁴ More surprising, it also failed to find a significant reduction in iron deficiency or anemia between groups, a much weaker hematologic signal than the pooled trials above showed. Whether that reflects this dose, this population's low baseline risk, or chance in a trial of 221 is not clear.⁴

A 2026 follow-up complicates the null, cautiously

The same Polish-Swedish cohort was checked again in 2026, three years after the original trial ended, this time for behavior rather than cognition. Of the original 221 infants, 133, or 60%, completed a standard behavior checklist at age three.¹⁰ The children who had received iron scored significantly lower on externalizing behavior (mean 45.6 vs. 48.6, adjusted P=.006) and on aggressive-behavior items (adjusted P=.033) than the children who had received placebo.¹⁰ Cognition, language, and motor scores in the same children showed no effect at any point through three years. The new result is specific to behavior.¹⁰

This is a pre-specified secondary outcome, not the trial's primary question. And 40% of the original cohort was lost to follow-up by three years, enough attrition that if the children who dropped out differed systematically from those who stayed, the result could shift in either direction.¹⁰ The trial's own authors call the finding a "potential beneficial effect" and say it "warrants cautious interpretation... until replicated."¹⁰ It is a genuinely new, credible finding that keeps the prevention question open rather than closing it. It is not evidence strong enough on its own to change a supplementation decision.

Where AAP and ESPGHAN part ways

The two major pediatric bodies on either side of the Atlantic draw different defaults from much of the same evidence.

Two guideline bodies, reading overlapping evidence, disagree about what a healthy breastfed infant should get by default. The AAP figure is a considered judgment made under real uncertainty.

The AAP itself updated its guidance in 2026, replacing the report that had stood since 2010, dated June 22 and published in the July 2026 issue of Pediatrics. It could not be read in full: the publisher blocked automated access on every attempted route, including the DOI itself, which redirects to a live but gated article page.¹² The screening ages and dose figures on the update, given next, are drawn from the Academy's own public summary and an independent guideline digest that cross-checks against it, not from the report's own evidence grading, which stays unconfirmed.¹²¹³ Screening age now depends on feeding pattern: 9 to 12 months for predominantly breastfed infants, 15 to 18 months for infants transitioning to cow's milk.¹³ The core supplementation dose for breastfed infants is unchanged at 1 mg/kg/day.¹³ Preterm infants get a different, higher regimen entirely, about 2 to 3 mg/kg per day starting at two weeks, reflecting that they missed some or all of the third-trimester iron transfer a term infant got.¹³ That dose is calculated against a specific infant's weight and gestational history by the clinician managing that infant.

Iron is also not a substance with no downside at any dose, in any setting. In Pemba, Zanzibar, a region with very high background malaria transmission, a randomized trial gave routine iron and folic acid to 24,076 children under three.¹⁴ Supplemented children were 12% more likely to die or be hospitalized than children given placebo, a result serious enough that the trial's monitoring board stopped it early.¹⁴ That finding does not describe a healthy term infant in a non-malarial, high-income household. But it is why the World Health Organization scopes its own supplementation recommendation specifically to settings where anemia prevalence in this age group already exceeds 40%, rather than making the same blanket recommendation the AAP does.¹⁵ What the Pemba trial rules out is the idea that iron is risk-free at any dose, in any population. That is why every guideline in this record treats dosing as a clinician's calculation, not a parent's estimate.

Home choices and clinician calls

For a healthy, average-risk, breastfed six-month-old, the record supports a specific default. Getting enough iron protects blood counts and future stores, worth doing on its own terms. It has not been shown to change measured development in an infant who was not already anemic.⁹⁴ Three groups are the exceptions. Infants born preterm or at low birth weight missed some or all of the third-trimester iron transfer and need a different, higher, clinician-calculated dose from early infancy.¹³ Infants exclusively breastfed well past six months without iron-rich complementary foods established are past the point where breast milk alone was ever enough.¹ And any infant with a diagnosed anemia is in the group where the randomized evidence for benefit is real, not theoretical.³ Each is a pediatrician's specific call for that child.

What is squarely a household choice is what goes on the spoon starting around six months. Meat, poultry, and fish carry iron a baby's gut absorbs efficiently, which is why the World Health Organization recommends an animal-source food daily from six months, while noting the evidence behind that specific recommendation is still graded low-certainty.¹⁶ Iron-fortified cereal, at the lower per-tablespoon figures noted above, helps build volume toward the daily target, but it is not as concentrated a source as meat on its own. The WHO is explicit that fortified food is not a guarantee by itself, especially for an infant already at higher risk.¹⁶

The other lever a household actually controls is showing up for the screening a pediatrician already recommends. That means a hemoglobin or iron-status check at the age the current AAP guidance sets for that infant's feeding pattern, not a test ordered on a hunch.¹³ That test is what turns iron deficiency from a population statistic into a specific number for one child. It is the only way to know which group that child is in: the well-established one this record can speak to with confidence, or the narrower one, already anemic, where the evidence for treatment is strongest.