(P07-041-25) Identification of Growth Trajectories in Early Life: A Comparative Study of Malawi and the United States

Objectives: Early-life growth patterns may be associated with health risks, but there are different statistical approaches to describe trajectories. We aimed to evaluate latent class growth modeling as a method to describe and compare growth among children in two settings with different nutritional and cultural contexts.

Methods: Anthropometric data were obtained from two cohorts in Malawi and the USA at birth, 6–9 mo, 8–12 mo, and 11–15 mo. For each sex, we determined growth trajectories using latent class trajectory modeling for weight and weight-for-age z-score (WAZ) in both cohorts, and length, length-for-age z-score (LAZ), and weight-for-length z-score (WLZ) in Malawi. Model structure was optimized over the number of classes, variance structures, and linearity structure. Model adequacy was assessed using posterior probability assignment (APPA > 70%), odds of correct classification (OCC > 5.0), and relative entropy (E_k > 0.5). Trajectory separation was evaluated using Degrees of Separation (DoS).

Results: We included 600 children in Malawi and 100 in the USA. Identified growth trajectory classes ranged from 2 (WAZ among females in the USA) to 5 (WAZ among females in Malawi), although some classes were small. All models met adequacy criteria, with DoS ranging from 0.04 to 34.78. Females had similar intercepts to males but demonstrated more varied weight-based trajectories, while length-based trajectories were comparable between the sexes. Compared to the USA, the Malawi cohort revealed more classes for weight and WAZ, except for WAZ among males. In Malawi, weight-based trajectories resulted in more classes than length-based trajectories. WAZ, LAZ, and WLZ showed flat or decelerated growth trends, with intercepts varying between classes, while length and weight classes generally had common intercepts but varying trajectories.

Conclusions: The variation in growth trajectories differed markedly across settings. Data-driven approaches for trajectory identification are feasible with smaller sample sizes, and rigorous model adequacy criteria can be satisfied. Future studies should consider clinical relevance and proportion of identified trajectories in predicting outcomes.

Funding Sources: Bill & Melinda Gates Foundation, Mongolia Mengniu Dairy (Group) Company Ltd, Indonesia Endowment Fund for Education (LPDP).