Diverse weaning foods and diet patterns at multiple time points during infancy period and their association with neurodevelopmental outcomes in 6-year-old children

Neurodevelopment proceeds rapidly during fetal life and infancy, a period of high brain plasticity during which environmental and nutritional inputs can have substantial effects. Prior work implicates factors such as parent–child interactions, environmental complexity, socioeconomic status, and nutrition in shaping developmental trajectories. An enriched environment—including adequate nutrition and varied sensory inputs—can promote neurogenesis and favorable neural circuit development, whereas restrictive or monotonous diets have been linked to early neurodevelopmental problems. Animal studies indicate that soft or limited diets may impair hippocampal cell proliferation, and human data suggest that lower dietary diversity in early childhood associates with delayed development. Given that infant diets change over time, analyses at a single time point may miss important patterns. The authors hypothesized that dietary patterns across multiple early-life time points influence neurodevelopment later in childhood. The study aimed to assess how diverse weaning foods and diet patterns from infancy to 3 years, derived via cluster analysis while accounting for socioeconomic and clinical covariates, relate to neurodevelopmental outcomes at age 6 across six K-DST domains.

The study situates its hypothesis within evidence that enriched environments bolster neurogenesis and cognitive development, while restricted diets correlate with neurodevelopmental problems. Animal models show detrimental effects of soft or limited diets on hippocampal neurogenesis. Human observational data from rural China found that lower dietary diversity and fewer food items consumed are associated with delayed early childhood development. Broader literature underscores nutrition’s role in neurodevelopment, with growth during infancy linked to later IQ, and micronutrients (iron, zinc, iodine) being critical during complementary feeding. Additional lines of research suggest diet influences the gut microbiome, which in turn modulates brain function and early cognitive outcomes, with distinct microbiome patterns reported in relation to neurodevelopmental profiles.

Design and data sources: Administrative, observational cohort using merged data from South Korea’s National Health Insurance Service (NHIS) and the National Health Screening Program for Infants and Children (NHSPIC). The NHIS provides demographics (birth date, sex, insurance premium, residence) and healthcare utilization (ICD-10 diagnoses, medications, procedures). NHSPIC comprises seven rounds from 4 to 72 months, including questionnaires, anthropometrics, physical exam, and the Korean Developmental Screening Test (K-DST). Ethics approvals were obtained; de-identified data obviated consent. Study population: From 2,395,966 children born 2008–2012, those completing NHSPIC rounds 1–4 with appropriate diet questionnaires (n=408,077) and K-DST at round 7 (n=714,364) were considered; 180,563 met inclusion. Exclusions included death; birthweight <2.5 kg or >4 kg; multiple or preterm birth; specified perinatal disorders or congenital anomalies; ICU admission >4 days before age 1; general anesthesia before age 1 or >5 days before age 2. Final eligible cohort: 133,243. Exposure: Parent-reported dietary information across NHSPIC rounds: Round 1 (4–6 months): milk type (breast, formula, mixed). Round 2 (9–12 months): timing of complementary food introduction, frequency, ingredients. Round 3 (18–24 months): frequency/amount of fruit juice or sweetened beverages. Round 4 (30–36 months): juice/sweetened beverage amount, meal frequency, milk intake. Dietary clustering: Polytomous Variable Latent Class Analysis (poLCA) modeled 2–10 latent clusters; model fit assessed via log-likelihood trends, BIC/AIC elbow heuristics, entropy (>0.6 indicating good separation), and minimum cluster size (>3% of participants). The optimal model yielded four clusters. Outcome: Neurodevelopment at 66–72 months via K-DST across six domains (gross motor, fine motor, cognition, language, sociality, self-care). Parent-completed items per domain; outcomes categorized as advanced, age-appropriate, need follow-up, or recommend further evaluation. Unfavorable outcome defined as need follow-up or recommend further evaluation in any domain or total score. Covariates: Sex; region at birth (Seoul, metropolitan, city, rural); economic status (health insurance premium quintiles); calendar birth year; birthweight; head circumference at 4–6 months; perinatal conditions (ICD-10 P-codes); atopic dermatitis and food allergy (per definitions in Supplementary Table 2). Statistical analysis: Descriptive statistics with chi-square and t-tests across clusters. Multivariate logistic regression estimated adjusted odds ratios (aORs) and 95% CIs for associations between dietary clusters and unfavorable K-DST outcomes (total and domain-specific), adjusting for all covariates. Interaction p-values between ORs were calculated using log-difference Z-tests. Analyses used R poLCA 1.6.0.1 (R 4.1.3) and SAS 9.4; two-sided P<0.05 indicated significance.

- Four dietary clusters among 133,243 children: Control (53.4%; n=71,169), Cluster 1 (36.0%; n=47,990), Cluster 2 (6.6%; n=8,750), Cluster 3 (4.0%; n=5,334). - Cluster characteristics: Control: highest human milk prevalence in infancy (51.6%), timely weaning (4–6 months; 85.8%), diverse ingredients by 1 year, low juice intake at 2–3 years, appropriate meal frequency and milk amount. Cluster 1: later weaning more common (>6 months; 23.6%), and selective/picky intake—lower fruit (53.3%), egg (22.5%), and fish (24.6%) at 1 year. Cluster 2: lowest human milk (22.6%), later weaning (35.8%), diverse ingredients but high juice intake at age 2–3 (≥200 ml/day: 23.5%–26.8%), fewer than three meals/day in many children by <3 years. Cluster 3: high exclusive human milk reporting (40.4%), latest weaning (36.8%), lowest ingredient diversity at 1 year (e.g., fish 9.7%, meat 12.6%, eggs 14.1%, fruits 26.1%, vegetables 41.6%, grains 49.9%). - Unfavorable total K-DST outcomes: Control 6.2% (n=4,448); Cluster 1 7.5% (n=3,596); Cluster 2 9.1% (n=794); Cluster 3 11.0% (n=586). - Adjusted odds ratios (aOR) for unfavorable total K-DST vs Control: Cluster 1: 1.209 (95% CI, 1.156–1.266); Cluster 2: 1.418 (1.312–1.532); Cluster 3: 1.741 (1.593–1.903); increasing gradient from Cluster 1 to 3 (interaction p-values <0.05). - Domain-specific aORs vs Control showed consistent patterns with stronger associations in Cluster 3: Gross motor: C1 1.186 (1.084–1.297); C2 1.300 (1.108–1.524); C3 2.030 (1.724–2.389). Fine motor: C1 1.279 (1.176–1.390); C2 1.535 (1.336–1.764); C3 2.248 (1.940–2.605). Cognition: C1 1.231 (1.143–1.326); C2 1.644 (1.459–1.851); C3 2.085 (1.825–2.382). Language: C1 1.355 (1.260–1.457); C2 1.634 (1.450–1.841); C3 2.270 (1.997–2.580). Sociality: C1 1.279 (1.176–1.391); C2 1.541 (1.338–1.775); C3 2.055 (1.762–2.396). Self-care: C1 1.345 (1.221–1.482); C2 1.482 (1.254–1.752); C3 2.111 (1.765–2.526). - Baseline differences: Cluster 2 had slightly more males; clusters varied by region and economic status (e.g., control more often higher quintiles; Cluster 3 more often lower quintiles). Mean birthweight ~3.22 kg across clusters; head circumference at 4–6 months slightly smaller in Clusters 1 and 3 than control.

Findings support the hypothesis that early-life dietary patterns, captured across multiple time points, are associated with neurodevelopment at age 6. Children whose feeding practices aligned with guidelines (breastfeeding, timely complementary feeding at 4–6 months, diverse ingredients, appropriate meal frequency, and moderate milk/low juice) had the lowest risk of unfavorable K-DST outcomes. Increasing deviation from these practices—picky/limited ingredient intake (Cluster 1), higher sugary beverage intake and fewer meals (Cluster 2), and markedly low dietary diversity with late weaning (Cluster 3)—was associated with a graded increase in risk across all developmental domains. Potential mechanisms include: (1) enriched environmental effects, where varied sensory and gustatory exposures from diverse complementary foods may promote neurogenesis and neural circuit development; (2) nutritional adequacy, as diverse weaning foods supply critical micronutrients (iron, zinc, iodine) and energy supporting growth and neurodevelopment; and (3) diet–microbiome–brain interactions, with early diet shaping gut microbiota that can influence cognitive and behavioral outcomes. The large, representative cohort and poLCA-based temporal dietary profiling strengthen the generalizability and the nuanced understanding of how multi-timepoint patterns, rather than single snapshots, relate to later neurodevelopment.

Dietary practices during infancy and early childhood—including feeding type, timing of complementary food introduction, ingredient diversity, meal frequency, and juice/milk consumption—are associated with neurodevelopmental outcomes at preschool age, with more guideline-concordant patterns linked to lower risk of unfavorable K-DST results. These findings underscore the importance of promoting timely, diverse complementary feeding and limiting sugary beverages in early life. Future research should incorporate prospective, detailed dietary assessments, repeated developmental evaluations, and broader environmental and familial factors to clarify causal pathways and to inform tailored dietary guidance for caregivers.

- Dietary exposure data were based on caregiver recall questionnaires, introducing potential recall bias, though surveys at multiple time points may mitigate some bias. - Neurodevelopment was assessed using the K-DST at a single time point (66–72 months), limiting the ability to capture trajectories; individual variability may lead to misclassification, as some children suspected of delay can normalize later. - Potential unmeasured confounding from educational attainment of children/parents and other growth environment factors (siblings, family, community) was not captured in the dataset. - Although extensive exclusions improved cohort homogeneity, residual confounding by health conditions or care practices may remain. - Cluster assignments and interpretations depend on observed variables; latent class models may not capture all nuances of diet quality or context.