Infant weight status in the first year of life offered little help in predicting whether health nurses would later record concerns about a child's development or well-being in the early school years, a cohort study of 54,917 children found. Across the indicators examined, AUC was no higher than 0.533. AUC, or area under the curve, is a score that indicates how well a model separates children with and without a recorded concern. The consistently low results left infant weight status with little value as a standalone predictor in this analysis.
Some category links stood out
The broad finding did not erase every category-level link. Severe thinness, thinness, overweight and obesity were associated with selected language or motor concerns, although the reported confidence intervals were wide. For language skills, severe thinness was associated with 1.82 times the odds of a recorded concern, with a 95% confidence interval from 1.08 to 2.91. Thinness was associated with 1.24 times the odds, with a 95% confidence interval from 1.04 to 1.47. These figures describe associations between weight categories and recorded concerns, not proof that one caused the other.
The researchers used a prospective cohort design, following children from an infant health-nurse visit to a school health examination. The final analytic group included 54,917 children after records with missing or unrealistic weight or recumbent-length measurements were excluded. The examinations covered the years 2007 to 2023.
Infant weight status was based on the health nurse's measurement and expressed as a WHO BMI z-score, which places a child's BMI on a common reference scale. At the school examination, nurses recorded concerns in several domains: language skills, motor development, perceived general well-being, perceived school well-being, the parent-child relationship, observed child interaction and peer relationships.
Across those recorded outcomes, nurses reported concerns about language skills for 9.8% of children and about motor development for 8.6%. The corresponding proportions were 5.1% for perceived general well-being, 10.6% for perceived school well-being, 5.0% for the parent-child relationship, 14.5% for observed child interaction and 7.4% for peer-relationship problems.
The formal prediction test stayed weak
To see whether the way weight was represented changed the result, the analysis used logistic regression with categorical weight status, a linear BMI z-score term and restricted cubic splines. Logistic regression estimates how an outcome varies across categories, while splines allow a relationship to bend rather than forcing it into a straight line. The researchers evaluated model performance with AUC, also known as the c-statistic.
The categorical models produced AUC values of at most 0.51 across all outcomes. Treating BMI z-score as a continuous measure found no linear relationship with development or well-being outcomes, and AUC values ranged from 0.50 to 0.51. The spline models produced the highest reported AUCs for language-skill concerns, at 0.514, and motor-development concerns, at 0.517. The values for the other outcomes were lower.
A result with clear limits
Extra checks did not materially alter the picture. Excluding children with early well-being concerns did not improve AUC. Analyses separated by sex showed no differences, while analyses grouped by parental education showed no clear pattern.
The authors' interpretation was that infant weight status, whether treated as categories or as a continuous BMI measure, was not a reliable standalone predictor of the later development and well-being concerns recorded by health nurses. The findings point toward testing additional or combined predictors if the aim is to forecast those outcomes.
There are important limits to the result. The models were not adjusted for additional confounders, external validation was not prioritized, and the full dataset was used for both model estimation and performance evaluation. The study therefore supports a cautious conclusion about this cohort and its health-nurse-recorded indicators, not a causal claim or a guarantee that the same model would work in another setting.
The study was funded by the Centre for Childhood Health, and the funding agency had no role in its design, analysis or writing.
Paper data and sources
Original title: Can weight status in the first year of life predict well-being in the first years of school?
Authors: Mie Sylow Liljendahl, Mette Rasmussen, Sofie Weber Pant et al.
Journal/Repository: International journal of obesity (2005)
Status: Corrected publication
First online: 2026-08-20
DOI: 10.1038/s41366-026-02180-6
Original paper · Full text