Dietary patterns and adverse pregnancy outcomes: Complementary insights from a priori and a posteriori approaches in...

Maternal diet during pregnancy may influence fetal growth, but evidence based on dietary-pattern approaches remains scarce. We characterized prenatal dietary patterns in a pregnancy cohort using two a priori diet-quality indices and an unsupervised a posteriori data-driven approach, and evaluated their associations with gestational and neonatal outcomes, and assessed the overlap between both methodological approaches.

We analyzed 595 mother-infant dyads from the Nutrition in Early Life and Asthma (NELA) birth cohort. Maternal diet at 19-22 weeks' gestation was assessed using a validated 123-item food frequency questionnaire. Adherence to the Mediterranean diet and to the Planetary Health diet was quantified using the relative Mediterranean Diet score (rMED) and the Planetary Health Diet Index (PHDI). In addition, we used principal components analysis (PCA) on energy-adjusted food groups, and then hierarchical clustering on the retained scores to identify real-world dietary patterns. Multivariable linear regression analysis was used to explore the association between dietary patterns and gestational weight gain (GWG), neonatal birth weight, and head circumference. A multinomial logistic regression model assessed the capacity of a priori indices to predict cluster membership, evaluated through McFadden's pseudo-R², overall accuracy, and Cohen's kappa.

Five dietary clusters were identified, with the following proportion of the population in each cluster: Cluster I (9.7%), Cluster II (31.1%), Cluster III (46.7%), Cluster IV (11.4%), and Cluster V (1.0%). Cluster II showed the highest adherence to both indices (mean rMED 9.4; mean PHDI 85.6). Cluster IV (second highest adherence) was consistently associated with lower GWG and higher birth weight compared with Cluster II in adjusted models (-1.28 kg; 95% CI -2.55 to -0.02 and 134.4 g; 95% CI 30.42 to 238.4, respectively) but not head circumference. No associations were found for the rest of clusters or rMED and PHDI indexes. The comparison between approaches revealed poor classification accuracy (55.5%), and slight-to-fair agreement (Cohen's κ = 0.24).

An empirically derived dietary profile characterized by vegetables, tubers, legumes, and dairy intake, was associated with lower GWG and higher birth weight. The limited agreement between approaches highlights that a priori indices and data-driven methods capture complementary dimensions of diet, supporting their combined use.