Identifying, Predicting, and Managing Second-generation Antipsychotics Associated Weight Gain (AAWG) in Children and Adolescents
Introduction: Children and adolescents prescribed second-generation antipsychotic (SGA) medications often experienced antipsychotic-associated weight gain (AAWG). Antipsychotic-associated weight gain (AAWG) can increase the risk of type 2 diabetes, hypertension, dyslipidemia, and cardiovascular disease. However, limited studies examined the longitudinal development of AAWG, and few studies have examined the feasibility of AAWG prevention and early intervention by classifying patients into different risk groups of AAWG. Despite weight loss being a well-understood adverse effect of psychostimulants, the most commonly prescribed concurrent medications for SGA in children and adolescents, existing data regarding the modification effect of AAWG are inconsistent. Examining these interconnected research questions is crucial for developing comprehensive strategies to address AAWG and optimize outcomes in this vulnerable population. Objectives: The overarching goal of this study was to comprehensively examine the risk of and the treatment for AAWG in children and adolescents prescribed second-generation antipsychotic (SGA) treatments. This overarching goal was pursued through three distinct study aims: 1) Identifying latent trajectories of AAWG among pediatric SGA recipients and assess associated risk factors. 2) Developing a prognostic-based machine learning algorithm for dynamic, real-time prediction of AAWG risk in individual pediatric patients. 3) Examining the effectiveness of concomitant psychostimulants in mitigating AAWG in a large cohort of children and adolescents undergoing SGA treatments. Methods: Utilizing the IQVIA Ambulatory EMR-US database (2016-2021), this study focuses on SGA-naïve patients aged 6–19 with at least 90 days of continuous SGA prescriptions. The assessment of weight change in children and adolescents utilizes BMI z-scores based on the CDC growth chart for ages 0 to 20. In Aim 1, the AAWG trajectories over a 24-month period were described using Group-based trajectory modeling (GBTM), and the risk factors associated with GBTM-identified trajectories were assessed using multinomial logistic regression. In Aim 2, six supervised machine-learning algorithms were developed to predict real-time AAWG and classify patients into subgroups at minor, moderate, and severe AAWG risks. In Aim 3, the effectiveness of SGA concomitant psychostimulants on AAWG was assessed using piecewise linear mixed-effects regression. Results: In Aim 1, the GBTM model identified that children and adolescents prescribed SGA followed 4 distinctive trajectories, including persistent severe (4.2%), persistent moderate (20.1%), minor change (69.6%), and gradual loss (6.1%). Risk factors associated with the membership of AAWG trajectories were younger age, lower baseline BMI z-score, and receiving olanzapine as the index SGA. In Aim 2, the predictive models demonstrated that the individual level AAWG risk could be accurately predicted using supervised machine learning algorithms, with the best-performed multiclass Xgboost model having an Area Under the Curve (AUC) at 0.921. Crucial predictive features were BMI z-score slope, baseline BMI z-score, SGA treatment duration, duration between the last BMI z-score measure and the BMI z-score score to be predicted, and counts of BMI z-score at follow-up. In Aim 3, the piecewise linear mixed-effects regression results suggested that psychostimulant was not associated with significant AAWG reduction as compared to the nonusers. Conclusions: The findings of the study demonstrated that children and adolescents prescribed SGA followed distinctive weight gain trajectories. The risk of AAWG can be accurately predicted using supervised machine learning algorithms. However, psychostimulants, the most used concurrent medication of SGA in children and adolescents, did not appear to have an AAWG reduction effect. The findings of our research have added to the literature and informed the feasibility of personalized AAWG monitoring and targeted intervention in children and adolescents prescribed SGA.