Endocrine Abstracts

Background: Recombinant human growth hormone (r-hGH) is the primary therapeutic agent for disorders of growth including growth hormone deficiency (GHD) and Turner syndrome (TS). There is a high cost associated with treatment and existing methods to predict response (and hence alter management) can only account for 40–60% of the variance.

Methods: GHD (n=71) and TS patients (n=43) were recruited as part of a study (PREDICT) on the long term response to r-hGH over five years of therapy¹. Change in height over the entire study and height velocity at each year of the study were used as endpoints to measure the effect of r-hGH. Pharmacogenomic analysis was performed using 1219 genetic (DNA) markers along with whole genome transcriptome (mRNA) from blood. Transcriptomic data were initially analysed using partial least square discriminant analysis (PLS-DA) to determine potential predictive value. Similarity in response to r-hGH between GHD and TS was assessed using gene interaction networks. Random forest, a machine learning technique, was used to define predictive value of gene expression data associated with growth response at each year of the study.

Results: No genetic marker passed the stringency criteria required for predictive value. Using PLS-DA and random forest we demonstrated that the transcriptomic data can be used to predict growth response to r-hGH at each of the five years and over the entire duration of the study in GHD and TS. Network models identified an identical core set of genes present in both GHD and TS at each year of therapy whose expression can be used to classify therapeutic response to r-hGH.

Conclusions: DNA markers are useful in growth prediction. However the transcriptome can be used to predict both short and long term therapeutic response to r-hGH. For the first time, core sets of genes identical in both TS and GHD patients can be used to predict response to r-hGH at each of the five years of the study.

Reference

1. Clayton P, Chatelain P, Tato L, et al. Eur J Endocrinol 2013; 169(3): 277-89.

Acknowledgement: PREDICT investigator group.