Group-Based Trajectory Modeling to Identify Patterns of Antipsychotic-Associated Weight Gain Among Children and Adolescents.

Purpose/background: Antipsychotic-associated weight gain (AAWG) is a common adverse effect of second-generation antipsychotic (SGA) medications among children and adolescents. This study applied group-based trajectory modeling to identify latent trajectories of AAWG among children and adolescents and associated risk factors.

Procedures: This was a retrospective analysis of the IQVIA Ambulatory EMR-US database from 2016 to 2021. The cohort consisted of patients aged 6 to 19 years who were SGA naive and received at least 90 days of continuous SGA prescriptions. Group-based trajectory modeling was used to identify latent trajectories of AAWG development during a 24-month period since SGA initiation, and multinomial logistic regression analysis was conducted to examine the risk factors associated with the identified AAWG trajectories.

Findings/results: A total of 16,262 patients were included. Group-based trajectory modeling identified the following 4 distinctive AAWG trajectories: persistent severe weight gain (4.2%), persistent moderate weight gain (20.1%), minor weight change (69.6%), and gradual weight loss (6.1%). Compared with the minor weight change group, younger age (12-17 vs 5-11: odds ratio [OR], 0.634; 95% confidence interval [CI], 0.521-0.771), lower baseline body mass index z -score (OR, 0.216; 95% CI, 0.198-0.236), and receiving olanzapine as the initial SGA (olanzapine vs aripiprazole: OR, 1.686; 95% CI, 1.673-1.699) were more likely to follow severe weight gain trajectories. The area under the receiver operating characteristic curves for comparing severe weight gain versus minor weight change groups and moderate weight vs minor weight change groups in the multinomial regression model were 0.91 and 0.8, respectively.

Implications/conclusions: A quarter of pediatric SGA recipients experienced persistent weight gain during the SGA treatment. The risk of having persistent AAWG can be predicted using patient characteristics collected before SGA initiation and the initial SGA agent.