ABSTRACT Background: Prenatal diet may contribute to metabolic alterations (e.g., insulin resistance) that could influence fetal programming and, subsequently, birth weight, a known determinant of impaired development, morbidity, and chronic disease risk such as obesity later in life1. Prenatal overall diet, as in a dietary pattern (DP), derived particularly in relevance to fasting plasma glucose, an intermediate variable between diet and birth weight, was found to be associated with birth weight2. These findings, however, are specific to predominantly non-Hispanic white women, limiting generalizability of findings to other populations such as Hispanic/Latina women who are at disproportionately greater risk of having adverse birth outcomes and children with childhood obesity3. Specific aims: 1) Derive prenatal DPs in a racially/ethnically diverse pooled sample from the two cohorts using RRR alternately with gestational weight gain (GWG) and fasting plasma glucose (FG) as intermediate response variables and characterize resulting DPs based on higher-loading foods; 2) Identify racial/ethnic differences in DPs by using the RRR approach in Aim 1 in groups stratified by race/ethnicity and compare racial/ethnic-specific DPs with DPs derived in the combined sample; 3) Determine whether racial/ethnic-specific prenatal DPs (Aim 2), versus those derived in a racially/ethnically diverse pooled sample (Aim 1), better predict birth weight and large-for-gestational age (LGA) by comparing how well they predict the outcomes within these racial/ethnic groups. Methods: Using existing data from two ECHO pregnancy cohorts, MyPyramid Equivalent food group daily intakes will be averaged across two or more Automated Self-Administered 24-hr dietary recalls and used as input variables in a reduced-rank regression model designed to derive linear combinations of foods that maximally explain the variability in two intermediate response variables (GWG and FG) hypothesized to be on the causal pathway between prenatal diet and the outcomes of interest (Aim 1). Additionally, DPs will be characterized based on higher-loading foods. The same approach from Aim 1 will be used in Aim 2 stratified by race/ethnicity (non-Hispanic white, Hispanic/Latina). DP differences by race/ethnicity will be based on higher food group factor loadings (>0.20) for similar food groups. In each racial/ethnic group, two sets of multivariable linear regression (birth weight) and logistic regression (LGA) adjusted for relevant covariates will be performed. The first set of models will include DPs derived from Aim 1 in each racial/ethnic group, while the second set will include DPs derived from Aim 2 in each racial/ethnic group. Qualitative comparisons between magnitude and strength of associations between sets of DPs in each racial/ethnic group will be performed to determine whether there are meaningful differences in prediction of birth weight between racial/ethnic-specific DPs and those derived from a m...