Interpretable machine learning identifies predictors of IGF-1 during growth hormone therapy in boys

Abstract

Insulin-like growth factor-1 (IGF-1) is widely used to guide growth hormone (GH) therapy, yet its predictors during the maintenance phase and puberty remain incompletely understood. We aimed to predict IGF-1 standard deviation scores (SDS) at 3 and 12 months in GH-treated boys and to identify key predictors of IGF-1 dynamics using interpretable modeling approaches. Data from 72 boys with growth hormone deficiency or idiopathic short stature from a previous Swedish clinical trial were analyzed using early-life characteristics, GH treatment variables, and puberty-related measures. Linear regression, Elastic Net regression, and explainable boosting machine (EBM) models were evaluated with nested five-fold patient-grouped cross-validation. Performance estimates were patient-balanced, with 95% confidence intervals based on 2,000 patient-cluster bootstrap resamples. Elastic Net had the highest point estimates at 3 months, but confidence intervals did not establish a difference between the primary models. No clear differences between models were found at 12 months. Current IGF-1 SDS was the most prominent predictive feature, while additional treatment- and puberty-related variables were also selected across cross-validation models. Interpretable modeling approaches provided moderate predictive performance for future IGF-1 SDS during the maintenance phase and puberty. The findings indicate that prediction is informed by current IGF-1 SDS together with additional clinical variables, but further validation in independent cohorts is required before clinical application.