Significant biomarkers for predicting 1-month changes in IGF-1 in growth hormone-deficient children following r-hGH therapy

Abstract

Growth hormone deficiency (GHD) is the most common pituitary hormone deficiency and is clinically characterized by short stature, delayed bone age and central distribution of body fat, and it has also been proven to be mildly heritable. Treatment with recombinant human growth hormone (r-hGH) is primary and safe for GHD children, and a dose of 0.15‒0.20 mg/kg each week results in a considerable increase in height velocity, with noteworthy growth during the first year of therapy [1]. Previous studies have shown that serum IGF-1 is strongly correlated with the growth response [2]. Therefore, IGF-1 can serve as a clinical indicator for monitoring compliance, efficacy and safety. However, the response to GH therapy shows significant individual variation, which is strongly associated with genetic factors. The prevalence rate of severe childhood GHD-related short stature varies from 1:4000 to 1:10,000 [3], while approximately 3%‒4% of the population in China suffers from short stature with an increasing trend. Therefore, an open-label, prospective, multicentric, noncomparative, nonrandomized phase IV interventional study (NCT01187550, Merck Serono Study 27709) was conducted to investigate the relationship between the prospective biomarkers of GHD patients and the individual variation in the primary therapeutic response following 4 weeks of r-hGH therapy. Given the significance of predicting GHD treatment response and the gaps in previous research, we sought to adopt a comprehensive strategy to accurately predict the therapeutic response utilizing the transcriptome, single nucleotide polymorphisms (SNPs) and clinical factors. We employed continuous variables and standard deviation scores of differences in serum IGF-1 levels after 4 weeks of r-hGH therapy (ΔIGF-1) as targets to filter possible influencing variables. Furthermore, we compared several potential machine learning techniques, validated by PCA and PLS-DA, and ultimately applied the elastic net algorithm to determine the optimized predictive factors with consistent effect sizes. Additionally, expression quantitative trait locus (eQTL) analysis and differentially expressed gene (DEG) analysis were conducted to identify significant biomarkers for GHD treatment.

1. Introduction

Growth hormone deficiency (GHD) is the most common pituitary hormone deficiency, characterized by short stature, delayed bone age, and central fat distribution, with a mild heritable component. Recombinant human growth hormone (r-hGH) therapy is the primary and safe treatment, typically administered at 0.15–0.20 mg/kg per week, leading to significant catch-up growth, especially in the first year. Serum IGF-1 levels strongly correlate with growth response, making it a valuable clinical biomarker for monitoring efficacy and safety. However, individual responses vary considerably, influenced by genetic factors. The prevalence of severe childhood GHD-related short stature ranges from 1:4000 to 1:10,000, with 3–4% of the Chinese population affected by short stature, a trend that is increasing.

To address the need for personalized treatment, an open-label, prospective, multicentric phase IV study (NCT01187550) was conducted to identify biomarkers predicting the therapeutic response after 4 weeks of r-hGH therapy. This study employed a comprehensive approach integrating transcriptomic data, single nucleotide polymorphisms (SNPs), and clinical factors. Using machine learning techniques, including elastic net, and validated by PCA and PLS-DA, we aimed to identify optimized predictive factors. Additionally, eQTL and differential gene expression analyses were performed to uncover significant biomarkers for GHD treatment response.