A skin sample taken early in life can help researchers ask what was measurable before a child was diagnosed with eczema. In Singapore, Ta and colleagues compared early skin measurements in infants who later differed in diagnosis. The timing makes this a useful study of possible early markers. The composition of the sample matters just as much when asking what a future prediction tool would tell parents.

THE SHORT READ
  • The nested analysis included 119 infants, 60 with atopic dermatitis diagnosed by 18 months and 59 controls.
  • The published abstract lists scheduled sampling at week one and months three, six and 18.
  • The case fraction was 50.4% of the analytic comparison. Estimating a baby’s future risk would require validation in a group that reflects the population where the tool would be used.
THE STUDY AT A GLANCEEarly-life skin signals and eczema in S-PRESTO · Ta et al.
Paper date
21 November 2026
Setting
Singapore S-PRESTO cohort
Analytic sample
119 infants
Primary endpoint
Physician-diagnosed atopic dermatitis by 18 months

Original publication: 21 Nov 2026 · The date above refers to this brief.

THE NUMBERS, IN CONTEXT

Who was included in the infant skin analysis?

infants

Atopic dermatitis cases60
Controls59
065
The two bars show the abstract’s comparison counts. The case fraction is 60 ÷ 119, or 50.4% of this analysis.
THE EXPERIMENT, EXPLAINED

From an early measurement to a prediction question

  1. Set the prediction age

    Use measurements available at that time.

  2. Define the later outcome

    Record diagnoses by a fixed age, here 18 months.

  3. Test in another group

    Compare predicted probabilities with observed outcomes in a new cohort.

Press-News guide to a future prediction validation. The last step describes a proposed test.

What was measured and when

The team collected material from the surface of skin without visible lesions using tape strips. The abstract describes repeated collection at week one and at three, six and 18 months. The investigators compared patterns in molecules from the skin and its resident microbes. They report some group differences before recorded clinical signs. Measurements taken after a diagnosis answer a different question from those taken before it. A prediction model for a particular age would need to use only the information available by that age.

Source 1 ↗

The fraction describes the comparison group

The original S-PRESTO study enrolled women before pregnancy and followed families into childhood. The new paper describes a nested analysis of 119 infants, with 60 cases and 59 controls. Dividing 60 by 119 gives 50.4%. That percentage describes who was in the analysis. A frequency estimate for the parent cohort would require all eligible infants, their follow-up and a record of how the comparison participants were selected. A similar distinction applies when an image study compares fixed numbers of cases and controls.

Source 1 ↗Source 2 ↗

From ranking risk to estimating a probability

A candidate marker can help rank one infant above another in an analysis. A probability needs another check. Among infants given roughly the same predicted chance, the observed proportion diagnosed by 18 months should be close to that chance in an independent group. Researchers call this calibration. A paper may report how well a model separates the two outcome groups before anyone knows how well the predicted probabilities transfer to an ordinary clinic population.

Source 1 ↗

Which follow-up would make the result easier to apply?

A follow-up could freeze the prediction model at a stated age and test it with newly recruited infants. Readers could then see which measurements were available before diagnosis, how many visits were missed, how well probabilities matched outcomes, and how errors varied across settings. The available abstract reports promising model performance in broad terms. This review can establish the design and sample counts from that abstract, while the numerical performance and validation remain outside its scope.

Source 1 ↗
TRACE THE EVIDENCE

Keep the study sample separate from the risk estimate

Publisher abstract and publication record for the accepted article. Checked the analytic group counts, sample timing and endpoint. Performance metrics and validation details were outside the available source scope.

01What does 60 of 119 mean?

What was observed
Sixty cases and 59 controls formed the nested comparison.
Where the conclusion stops
The calculated 50.4% describes this analysis and requires a different sampling frame to estimate population frequency.

Source 1 · Published abstract

02What can the order of measurements show?

What was observed
Early skin signals were reported before clinical signs.
Where the conclusion stops
Timing helps motivate a prediction study. It cannot by itself identify a cause.

Source 1 · Published abstract, sampling and endpoints

Numbers you can inspect

MeasureValue & unitOrigin & method
Infants in the analysis119 infantsReported Nested comparison in the Singapore cohort.
Source 1 · Published abstract
Atopic dermatitis cases60 infantsReported Cases in the analysis with diagnosis by 18 months.
Source 1 · Published abstract
Controls59 infantsReported Comparison group in the analysis.
Source 1 · Published abstract
Case fraction of the analytic sample50.4 percentCalculated Divide 60 by 119 and multiply by 100, rounded to one decimal.
Source 1 · Published abstract
Age limit for the diagnosis endpoint18 months of ageReported Physician diagnosis assessed by this age.
Source 1 · Published abstract

Compare the actual experiments

These studies answer different questions. Read the unit and endpoint before comparing results.

StudyUnit & settingReadoutInterpretation boundary
Parent cohort and nested analysis

Source 2 · Loo et al., Study design

Families enrolled before pregnancy and 119 selected infantsProspective follow-up and a comparison of two outcome groupsAn outcome fraction in the nested analysis describes its composition.
Early marker and later prediction check

Source 1 · Published abstract, sample timing and endpoint

Infant skin samples and diagnoses in a separate groupCandidate marker timing and eventual calibrationAn independent group can show whether probabilities match diagnoses outside the original analysis.
Download evidence table (CSV)

The export includes claims, available numbers, methods and source locations. It contains our reading notes and published summaries; it is not raw participant data or an independent reanalysis.

Evidence update · 22 Nov 2026
First publication. Any calculations in this brief use reported figures. The reading notes identify possible follow-up questions. AI source check; no human editorial or clinical review.

CONNECT THE EVIDENCE

What three study details can tell a reader

DetailPublished abstractInterpretation
Early sampleSkin measurements starting in week onePossible lead time before a recorded diagnosis.
Cases and controls60 diagnosed infants and 59 controlsComposition of the analytic comparison.
Diagnosis windowPhysician diagnosis by 18 monthsThe outcome and time horizon a model would have to predict.

Press-News reading aid based on the primary abstract and parent cohort profile.

READER QUESTIONS

Your questions, answered

Did half of all babies in the cohort develop eczema?

The 50.4% figure comes from the counts in the nested comparison. To estimate frequency, use the whole eligible cohort and account for follow-up.

What condition did the researchers count?

The primary endpoint was atopic dermatitis, a form of eczema, diagnosed by a physician by the age of 18 months.

Is 50.4% a measure of model accuracy?

It is the proportion of cases in the analysed sample. The abstract does not give a numerical model performance estimate.

What is calibration in this example?

It is agreement between assigned probabilities and observed diagnosis frequencies at a stated age in the population where the model is evaluated.

LIMITATIONS

Limits of this interpretation

  • The publisher made an accepted article record and abstract available for this review. The abstract gives no numerical model performance estimates.
  • The reported group counts describe a nested analysis. The parent cohort’s diagnosis frequency requires a separate denominator.
  • A temporal lead can help identify candidate markers. A causal explanation would require other evidence.
SOURCE NOTES

Sources & transparency

  1. Ta, Wearne, Lim et al. (2026) — Early-life skin molecular signatures are predictive of atopic dermatitis risk in an Asian preconception birth cohort

    Publisher abstract and publication record for the accepted article. Checked the analytic group counts, sample timing and endpoint. Performance metrics and validation details were outside the available source scope. · Accessed 22 Nov 2026

    DOI: 10.1038/s41467-026-79674-4
  2. Loo et al. (2021) — Cohort profile of S-PRESTO

    Published profile of the parent cohort. Checked enrolment before pregnancy and the design for repeated infant visits. · Accessed 22 Nov 2026

    DOI: 10.1007/s10654-020-00697-2

Prepared and source-checked with AI for Press-news Team. The source notes identify the material available during preparation. The article received no human editorial or clinical review. The source list supports the reported results and the calculations shown.

Source check: AI source check — primary research and stated source limits

Clinical review: Not applicable to this educational guide

Suggest a correction