Only about 3% of children follow the 'atopic march' from eczema to asthma to hay fever

The "atopic march" describes a sequence: eczema in infancy, then food allergy, then asthma, then hay fever, as if allergic disease progressed along a fixed track. When two population-based birth cohorts totalling 9,801 children were analysed for the actual patterns their symptoms followed, that sequence turned out to describe 3.1% of them — around 7% of the children who had any symptoms at all [s1]. It is a real pattern. It is not the usual one.

The finding matters because the march is routinely used to tell parents of a baby with eczema what is coming next.

How the analysis was done

Children in the two cohorts were assessed at ages 1, 3, 5, 8 and 11 years for eczema, wheeze and rhinitis [s1]. Rather than testing whether the group-level prevalence of each condition rose and fell in the expected order — which is how the atopic march was originally described — the researchers used Bayesian machine learning to identify latent classes based on each individual child's profile over time [s1].

That distinction is the study's argument. The authors propose that the atopic march paradigm arose from cross-sectional analyses of longitudinal studies, and may describe a population pattern that does not predominate at the individual level [s1]. A population in which eczema peaks early and rhinitis peaks late will look like a march in aggregate even if no individual child follows it.

The eight profiles

The data were best described by eight latent classes: no disease (51.3%), atopic march (3.1%), persistent eczema and wheeze (2.7%), persistent eczema with later-onset rhinitis (4.7%), persistent wheeze with later-onset rhinitis (5.7%), transient wheeze (7.7%), eczema only (15.3%) and rhinitis only (9.6%) [s1]. When the modelling was run separately for each cohort, the results were similar [s1].

Patterns of allergic sensitisation were highly concordant but were associated with different profiles of eczema, rhinitis and wheeze [s1] — meaning sensitisation does not sort children into these tracks either.

The authors' conclusion is that the developmental profiles are heterogeneous, and that only a small proportion of children, about 7% of those with symptoms, follow trajectories resembling the atopic march [s1]. The main limitation they report is a difference in the wording used to ascertain eczema, wheeze and rhinitis between the two cohorts [s1].

What infant eczema does predict

Rejecting the march as a general rule is not the same as saying infant eczema carries no information. The most direct evidence that it does comes from the trial design of LEAP, which needed to enrol infants at high risk of peanut allergy and used severe eczema, egg allergy, or both, as the entry criterion [s2]. Among the 530 trial infants with a negative skin-prick test at entry who were assigned to avoid peanut — the standard advice at the time — 13.7% were peanut-allergic at 60 months [s2]. Among the 98 who already had a positive skin-prick test and were assigned to avoidance, 35.3% were [s2].

Those rates are far above general-population peanut allergy prevalence, which is the point: severe infant eczema selected a group whose peanut allergy risk was high enough to run a prevention trial in.

A mechanism that links skin to food

There is also a genetic thread connecting the two. Loss-of-function mutations in the filaggrin gene are associated with atopic dermatitis and other atopic diseases, which made filaggrin a candidate gene in the causation of peanut allergy [s3]. A case-control study tested whether they are also associated with peanut allergy: 71 English, Dutch and Irish patients with food-challenge-positive peanut allergy against 1,000 non-peanut-sensitised English population controls, with replication in 390 white Canadian patients and 891 Canadian controls [s3].

Filaggrin loss-of-function mutations were strongly associated with peanut allergy in the challenge-positive patients (P = 3.0 x 10⁻⁶; odds ratio 5.3, 95% CI 2.8 to 10.2), and the association replicated in the Canadian sample (P = 5.4 x 10⁻⁵; odds ratio 1.9, 1.4 to 2.6) [s3]. Crucially, it remained significant (P = .0008) after controlling for coexistent atopic dermatitis [s3] — so the shared genetics are not simply a proxy for having eczema.

The authors read this as evidence for a role of epithelial barrier dysfunction in the pathogenesis of peanut allergy [s3]. The case-control design and the modest sample sizes limit it, and the two odds ratios differ substantially between populations, which is common in genetic association studies and a reason not to treat the larger figure as the effect size.

What this changes

Two claims are often merged. "Infant eczema is a marker of elevated allergy risk" is well supported: it was the criterion used to define a high-risk trial population [s2], and there is a candidate mechanism connecting barrier dysfunction to food sensitisation [s3]. "Infant eczema means a child will proceed through asthma to hay fever" is not: the trajectory analysis puts that specific sequence at 3.1% of children [s1].

What remains unknown is who the 3.1% are before they march. Nothing in this literature identifies a marker that separates the child with eczema who will follow the sequence from the 15.3% whose eczema is the only thing that happens [s1]. Until it does, the march is a description of a minority pattern rather than a prognosis.

Sources

• Developmental profiles of eczema, wheeze, and rhinitis: two population-based birth cohort studies — PLOS Medicine, 2014-10-21
• Randomized Trial of Peanut Consumption in Infants at Risk for Peanut Allergy — New England Journal of Medicine, 2015-02-23
• Loss-of-function variants in the filaggrin gene are a significant risk factor for peanut allergy — Journal of Allergy and Clinical Immunology, 2011-03-07