Early physical growth and neurodevelopment in full-term small for gestational age infants: reference points from propensity score matching and restricted cubic splines

Background: Full-term small for gestational age (SGA) infants are at risk for impaired neurodevelopment, but quantitative descriptions of growth-related developmental patterns and heterogeneity within this population remain limited. This study aimed to examine the associations of 1-year neurodevelopment with physical growth and birth-related factors in full-term SGA and appropriate for gestational age (AGA) infants, and to explore heterogeneity within the SGA group using age- and sex-adjusted relative length at 1 year. We also sought to assess whether SGA status remained associated with developmental quotient (DQ) scores after accounting for measured covariates and estimate model-derived weight-ratio reference points corresponding to the average DQ levels of AGA infants.

Methods: In this retrospective cross-sectional study, 448 SGA and 2,376 AGA children were included. Birth data, anthropometric measurements (weight, length, and head circumference) at 1 year, and Gesell Developmental Scale (GDS) scores were collected. Generalized linear models (GLMs) were used to examine factors associated with DQ in the SGA and AGA groups separately. Heterogeneity within the SGA group was further explored by stratifying infants according to age- and sex-adjusted relative length at 1 year. Two-step propensity score matching (PSM) with weighted GLMs was used to compare DQ scores between SGA and AGA infants after matching the measured covariates. Restricted cubic spline (RCS) models were applied to assess nonlinear associations between the 1-year weight ratio (SGA/AGA) and DQ.

Results: In the SGA group, 1-year weight was positively associated with DQ across multiple domains, including fine motor, adaptive behavior, language, and personal-social behavior (all P<0.01). Across age- and sex-adjusted relative-length strata, lower relative length was accompanied by lower weight, shorter length, smaller head circumference, and lower DQ scores across all five GDS domains (all P≤0.002). After matching the measured covariates, the SGA and AGA groups showed no statistically significant differences in DQ scores across domains (all P>0.05). Weight ratio showed nonlinear associations with DQ in all five domains (all P<0.001). The model-derived reference points at which the predicted mean DQ was closest to the average AGA DQ for the gross motor, fine motor, adaptive behavior, and language domains were 0.901, 0.909, 0.864, and 0.890, respectively. The personal-social domain showed a U-shaped pattern, with the lower crossing point at 0.741.

Conclusions: In this clinic-based retrospective cross-sectional sample, neurodevelopmental scores at 1 year in full-term SGA infants were more closely associated with postnatal growth status than with SGA status after matching. Relative length at 1 year also revealed substantial heterogeneity within the SGA group. The weight-ratio values identified by spline models may serve as statistical reference points for developmental comparison; however, they should not be interpreted as causal or universal clinical targets. Because development was assessed only at 1 year, these findings should be interpreted cautiously.