Longitudinal assessment of abdominal fat compartments in children may help delineate some of the early risk factors that can predispose an individual to high abdominal adiposity and insulin resistance. While accurate determination of abdominal fat can be achieved using computed tomography (CT) or magnetic resonance imaging (MRI), it is often not performed in large cohort studies involving young children due to radiation exposure, high costs or poor compliance with scan procedures. The goal of this work is to develop and validate predictive equations for abdominal fat compartments (subcutaneous (SAT) and internal (IAT) adipose tissue) from anthropometric values with MRI-based abdominal fat volumes as reference in multi-ethnic cohort of 4.5 year-old Asian children.Our data for the study included observations from 54 children (31 Chinese, 14 Malays and 9 Indians) aged 4.5 years from the Growing Up in Singapore Towards healthy Outcomes (GUSTO) study [1]. Anthropometric measures including height, weight, abdominal and upper mid-arm circumference and skinfold thickness (subscapular, triceps, biceps and suprailiac) were measured. Axial abdominal images with 5 mm slice thickness and in-plane resolution of 0.94 × 0.94 mm were acquired using water suppressed HASTE sequence (TR=1000 ms, TE=95 ms) on the 3T MR scanner (Siemens Skyra). The SAT and IAT volumes between the diaphragm and S1 sacral level were segmented with an automated algorithm [2] and edited manually for correction of any mis-segmented regions. The total sample (n=54) was randomly divided into training (n=30) and testing (n=24) groups. Statistical analyses were performed using SPSS 23.0 software. Linear predictive models were created for the SAT and IAT depots with anthropometric measures, ethnicity, weight, height and gender as potential predictors. Forward stepwise selection was performed using the Akaike information criterion (AIC) to select/remove predictors from the model in an automated fashion. The predictive equation derived from the training set was validated against the MRI-based fat volumes of the test set.The optimal predictive equations yielded by the linear modeling for SAT and IAT are shown in Table 1. Within the training dataset (Fig. 1 & 2), the equations had a strong predictive power for SAT (R² = 0.97) and a moderate predictive power for IAT (R² = 0.73). When the equations were cross-validated on the test dataset (Fig. 3 & 4), the R² for prediction of SAT was 0.88, while R² for IAT was 0.32. Body weight was the strongest predictor of IAT.We have developed equations to predict abdominal fat depot volumes using anthropometric measures in 4.5 year-old Asian children. We found that SAT can be predicted with high accuracy from anthropometric measures while equation for IAT had poor predictive power. These equations will benefit large cohort studies in Asian preschoolers for the interpretation of anthropometric data with regard to abdominal fat partitioning.