OR21-01 Pre-treatment Blood Transcriptome Predicts First-year Growth And IGF-I Response To Somapacitan Treatment In Children With GH Deficiency

Abstract

Abstract Disclosure: T. Garner: Grant Recipient; Self; Novo Nordisk. P.E. Clayton: Grant Recipient; Self; Novo Nordisk. M. Hojby: Employee; Self; Novo Nordisk. Stock Owner; Self; Novo Nordisk. P. Murray: Grant Recipient; Self; Novo Nordisk. A. Stevens: Grant Recipient; Self; Novo Nordisk. Background: Growth hormone (GH) replacement therapy often requires daily subcutaneous injections that can be burdensome for patients and their caregivers. Somapacitan is a long-acting, once-weekly GH derivative in development for treating GH deficiency (GHD) in children. Somapacitan reduces treatment burden and shows a similar efficacy and safety profile as daily GH in children with GHD. Here, we investigate the prediction of first-year growth and insulin-like growth factor-I (IGF-I) response based on baseline blood transcriptome in children with GHD treated with somapacitan or daily GH. Methods: 200 treatment-naïve prepubertal children with GHD were enrolled into REAL4, a randomised, multinational, open-labelled, and active-controlled phase 3 trial. 128 consented to a baseline blood transcriptome profile. Children were treated with once-weekly somapacitan (n=81) or daily GH (n=47) for 52 weeks and categorised based on treatment response: the upper quartile of height velocity (HV; cm/year) was defined as good responders, and the lower quartile as poor responders. We split the somapacitan group into training (70%) and testing (30%) sets for prediction validation of HV and the GH response marker: ΔIGF-I standard deviation score (SDS). Gene expression was assessed between the target quartile and the remaining quartiles for both HV and ΔIGF-I SDS to identify the top 100 differentially expressed genes by adjusted p-value. Classes were balanced using a synthetic minority oversampling technique and Boruta, a feature selection algorithm, was used to refine gene lists. We performed random forest and calculated “out of box” (OOB) area under the curve (AUC) and OOB error rate (ER), measures of predictive accuracy and robustness, respectively. Results: HV prediction from the transcriptome was excellent for both treatments (AUC range: 0.95-0.99). It was more robust for identifying good responders (ER: 7.0-7.5%) than poor responders (ER: 13.2-17.5%). Genes previously identified as predictive of growth response for daily GH (Stevens et al., 2021. The Pharmacogenomics J., 21:594-607) were demonstrated to have high predictive value for once-weekly somapacitan and daily GH treatment in this study (AUC range: 0.84-0.97, ER <22%). Using the somapacitan group for validation, HV prediction (AUC range: 0.78-0.81) was stronger than ΔIGF-I SDS prediction (AUC range: 0.63-0.72). Conclusions: We demonstrate pre-treatment blood transcriptome predicts first-year growth response for somapacitan. This represents an independent validation of earlier observations and shows that genes predictive of growth response for daily GH treatment in children with GHD are also predictive for once-weekly somapacitan. Further, we demonstrate that the pre-treatment blood transcriptome predicts growth response better than it predicts IGF-I SDS response. Presentation: Saturday, June 17, 2023