ESPE Abstracts

Background: `(LAGHs) are under investigation to treat SGA. Results from the REAL5 phase 2 trial (randomised, multinational, open-label; NCT03878446) indicate that the LAGH somapacitan has similar efficacy, safety, and tolerability profile as daily GH1. Predicting GH therapy response is critical to improve clinical management of short stature. Here, we compare the prediction of growth response in SGA children treated with daily GH or somapacitan based on the baseline blood transcriptome and clinical variables in the REAL5 trial.

Methods: Three somapacitan doses (0.16, 0.20, & 0.24 mg/kg/week; n = 12, n = 13, & n = 12) versus two daily GH doses (0.035 & 0.067 mg/kg/day; n = 12 & n = 13) were used. Because of limited sample sizes, dosage groups were merged. Four metrics for 52–week GH response were explored: height velocity (HV; cm/year), HV standard deviation score (HVSDS), change in height SDS (ΔHSDS), and change in IGF-I SDS (ΔIGF-I SDS). Participants were split into tertiles to define slowest responders and differentially expressed genes (DEGs) were identified from the blood transcriptome compared to remaining tertiles. Random forest (RF) models were created using the top 100 DEGs alone, 100 DEGs plus a set of clinical variables, or using clinical variables exclusively. RF prediction accuracy was presented as area under the curve (AUC) accompanied by error rates (ER) with subsequent validation performed. Accuracies were calculated for slow responders in each treatment.

Results: The ability of the transcriptome to identify SGA children who respond slowest in HV following treatment was highly accurate for both daily GH or somapacitan (AUC [ER))= 1.0 [5.0%) & 0.99 [6.9%), respectively). Prediction of other response measures was more variable (e.g. AUCs=0.88–0.99, ER HVDS = 6.9%, ER ΔIGF-I SDS = 7.7%, ER ΔHSDS = 13.8% [somapacitan]). Combining the transcriptome with clinical variables led to decreased accuracies and higher error rates for HV (0.90 [19%] for somapacitan). By comparison, clinical variables alone also performed poorly (0.82 [26%] for somapacitan).

Conclusions: This first use of transcriptome data demonstrated the capacity to identify poor responders in children born with SGA treated with daily or weekly GH. However, combining the transcriptome with adding clinical data or using clinical variables alone diminished performance. The baseline blood transcriptome in SGA has the potential for increased predictive power in comparison to clinical variables. Juul et al. 2023. Eur J Endocrinol, 188:1–12.