Predicting Weight Outcomes From Obesity Medications in a Paediatric Population

Area of Science:

• Pediatric Endocrinology
• Pharmacotherapy
• Data Science

Background:

• Limited real-world studies compare pediatric obesity medication (OM) outcomes.
• Understanding real-world variability and predictive factors is crucial for effective treatment.

Purpose of the Study:

• To analyze real-world outcomes of different OMs in adolescents.
• To develop predictive models for OM treatment success based on patient and treatment factors.

Main Methods:

• Utilized electronic health record data from 595 adolescents (12-21 years) with obesity.
• Assessed %BMI percentile above the 95th percentile (%BMIp95) changes.
• Employed Catboost machine learning to predict >=5% reduction in %BMIp95 using patient and treatment variables.

Main Results:

• OMs (Liraglutide, Phentermine, Phentermine/Topiramate, Semaglutide) showed comparable outcomes over 12 months.
• Semaglutide demonstrated a slightly higher average %BMIp95 reduction (12.87%).
• Predictive models achieved AUROC >= 0.75, with OM duration, baseline %BMIp95, and specialty visits as key predictors.

Conclusions:

• OMs studied had similar real-world effectiveness in adolescents.
• Predictive models can aid clinical decision-making for personalized OM therapy.
• Prospective validation of predictive models is recommended for future research.