Feeding high-risk babies peanut early prevents the allergy

How the advice reversed

Until 2008, the official advice ran the other way. Clinical guidelines in the United Kingdom in 1998 and the United States in 2000 told families of infants at high risk for allergy to keep peanut and other allergenic foods out of the diet, and the American version set the cutoff at three years of age.¹² The reasoning was that avoiding a food an infant might react to would keep the allergy from forming. It was intuitive, and it was wrong. A randomized controlled trial published in 2015, Learning Early About Peanut Allergy (LEAP), tested the opposite approach in exactly the babies the old rule targeted and found that feeding them peanut from infancy cut their rate of peanut allergy at age five by most of its value.¹ The guidance inverted within two years, from delay to early introduction.

The trial that carries the strong claim enrolled only high-risk babies, and the size of the benefit depends on how likely a child was to become allergic in the first place. The evidence is unusually clean for a nutrition question. It is cleanest exactly where the population is narrowest.

LEAP's two arms

LEAP randomly assigned 640 high-risk infants, between 4 and 11 months old, to either consume or avoid peanut until 60 months of age. The researchers split them by a baseline skin-prick test: one cohort with no measurable reaction to peanut, and a smaller one that already showed a small wheal.¹

In the larger cohort, the 542 infants with a negative baseline test, the result was stark. At age five, peanut allergy stood at 13.7% in the avoidance group and 1.9% in the consumption group. The absolute difference of 11.8 percentage points (95% confidence interval 3.4 to 20.3, P<0.001) is an 86.1% relative reduction.¹ Put in whole babies: of 100 high-risk infants who avoided peanut, about 14 were allergic at five; of 100 who ate it early, about 2. In the smaller cohort of 98 infants already sensitized to peanut, allergy ran 35.3% against 10.6%, a 70.0% relative reduction.¹ This is randomized evidence, a large effect, in high-risk infants only.

A relative figure alone tells a parent nothing to act on, so convert it. The number needed to treat is the count of infants who must get the intervention to prevent one case. LEAP's absolute reduction of 11.8 percentage points works out to about eight to nine high-risk infants fed peanut early for each peanut allergy prevented (100 divided by 11.8 is roughly 8.5).¹ For a high-risk baby, that is a large, cheap effect.

LEAP did not test a single taste. Infants in the consumption group ate at least 6 grams of peanut protein a week, spread across three or more meals, every week to age five. The median they actually managed was 7.7 grams a week, against 0 in the avoidance group.¹ A follow-up, LEAP-On, then asked every child to stop peanut for a year. At 72 months the protection held: 18.6% allergic in the original avoidance group against 4.8% in the original consumption group, with only three new cases in each group during the year off.³ Protection survived a pause, but it did so after five years of eating, which is not a license to introduce peanut once and stop. Safety was reassuring throughout, with no significant difference in serious adverse events between the groups.¹

EAT and the adherence trap

If LEAP answers the high-risk question, the Enquiring About Tolerance (EAT) trial was built to answer the general-population one. EAT recruited 1,303 exclusively breastfed infants from the general population at three months and randomized them to early introduction of six allergenic foods (peanut, cooked egg, cow's milk, sesame, whitefish, and wheat) or to standard UK practice of breastfeeding to around six months.⁴

The pre-specified analysis, by intention to treat, found nothing. Food allergy to at least one of the six foods developed in 5.6% of the early-introduction group and 7.1% of the standard group, a difference that could easily be chance (P=0.32, relative risk 0.80, 95% confidence interval 0.51 to 1.25).⁴ The per-protocol analysis, restricted to families who actually followed the feeding schedule, told a different story: any food allergy at 2.4% against 7.3% (P=0.01), and for peanut, zero cases among 310 adherent early-introduction infants against 13 cases among 525 in the standard group (P=0.003).⁴

The gap between those two analyses is about adherence, not about efficacy. Only 42.8% of the early-introduction group managed the demanding schedule of six foods a week in a young infant.⁴ A per-protocol result keeps only that 42.8% and drops the rest, and the families who can sustain such a regimen differ systematically from those who cannot: nonwhite race and parentally perceived reactions to the foods were among the strongest predictors of dropping out. So the per-protocol numbers show that early feeding can work when it is actually done. They cannot show that advising it to the general population will work at the population level, because the advice would reach the families who could not comply too. The authors themselves declined to claim efficacy and reframed the finding as a question about whether the effect is dose-dependent, while noting the regimen was safe but not easily achieved.⁴

Where the evidence thins

Pooling the trials sharpens the picture. A 2016 systematic review in JAMA estimated that early introduction was associated with a relative risk of peanut allergy of 0.29 (95% confidence interval 0.11 to 0.74) and of egg allergy of 0.56 (0.36 to 0.87), each rated moderate-certainty on the GRADE scale.⁵ In absolute terms that is roughly 18 fewer peanut cases and 24 fewer egg cases per 1,000 infants.⁵

Run the same arithmetic that made LEAP look so strong, and the population dependence becomes visible. LEAP's 11.8-point absolute drop gave a number needed to treat near 8. The pooled peanut estimate of about 18 fewer cases per 1,000 implies a number needed to treat near 56 (1,000 divided by 18). It is the same intervention with roughly seven times as many infants fed per allergy prevented, because an average baby's baseline risk is a fraction of a high-risk baby's, and a relative reduction applied to a small risk removes a small absolute number. The peanut pooled estimate is itself driven mainly by LEAP, so the general-population case rests on extrapolation and moderate-certainty pooling, not on a positive general-population randomized trial.⁵

The guidelines, tier by tier

The 2017 addendum from a US expert panel, sponsored by the National Institute of Allergy and Infectious Diseases (NIAID), translated LEAP into three risk tiers.⁶ Infants with severe eczema, egg allergy, or both should have peanut-containing food introduced as early as 4 to 6 months, with peanut testing strongly considered first; this tier carries moderate-quality evidence, because it maps directly onto LEAP. Infants with mild-to- moderate eczema should start around 6 months, at home, without testing, on evidence the panel graded low, since it extrapolates from a single study. Infants with no eczema or food allergy get peanut freely, in keeping with family and cultural preferences.⁶

The 2021 consensus from the American and Canadian allergy societies went further and simplified. It advises introducing both peanut and egg around 6 months, though not before 4 months, for every infant regardless of risk, and states plainly that screening before introduction is not required, that regular ingestion should be maintained once peanut is in, and that other allergenic foods should not be deliberately delayed either.⁷

On one point the two documents genuinely disagree. NIAID strongly recommends peanut testing before introduction for the highest-risk tier; the 2021 consensus says screening is not required for any infant and warns it can needlessly delay introduction.⁶⁷ Both are current and authoritative, so this is an evolution in progress. The safe way through it is the one both camps would accept: a genuinely high-risk infant, with severe eczema or a known egg allergy, is a conversation with the pediatrician before the first taste, not a decision to make alone.

What a reasonable parent does

The evidence changes the default, and the default is now early.

Introduce peanut and egg once solids begin, around 4 to 6 months, when the baby is tolerating first foods.⁶⁷ Keep them in the diet after that. Introduction without regular exposure does little; LEAP fed roughly 6 grams of peanut protein a week for years, and the guidelines ask for regular ingestion a few times a week.¹² Trying peanut butter or eggs once is not enough; the food has to keep coming back.⁸ For a baby with severe eczema or a known egg allergy, talk to the pediatrician first, since the testing guidance is exactly where the experts still differ.⁶

Then the safety floor, which is not optional. Never whole nuts before age five, and never peanut butter off a spoon or in lumps before age four, both of which are choking hazards. The vehicle is thinned peanut butter, peanut puffs, or peanut flour stirred into food the baby already eats.⁶ The strong number, the ~80% reduction, belongs to high-risk infants; for an average baby the case is reasonable rather than dramatic, and early introduction carries little downside either way.¹ Start early. Keep it up. Choking-safe forms only. Ask the pediatrician if the eczema is severe or the egg allergy already there.