ESPE Abstracts (2015) 84 P-1-83

Genetic Markers Contribute to the PREDICTION of Response to GH in Severe but not Mild GH Deficiency

Adam Stevensa, Philip Murraya, Jerome Wojcikb, John Raelsonc, Ekaterina Koledovad, Pierre Chatelaine & Peter Claytona


aFaculty of Medical and Human Sciences, Institute of Human Development, University of Manchester and Manchester Academic Health Science Centre, Royal Manchester Children’s Hospital, Manchester, UK; bQuartz Bio, Geneva, Switzerland; cGenizon BioSciences, St Laurent, Quebec, Canada; dMerck Serono, Darmstadt, Germany; eDepartment Pediatrie, Hôpital Mère-Enfant – Université Claude Bernard, Lyon, France


Background: Single nucleotide polymorphisms (SNPs) associated with the response to GH therapy have previously been identified in growth hormone deficient (GHD) children in the PREDICT long-term follow-up (LTFU) study (NCT00699855).

Objective and hypotheses: To assess the effect of GHD severity on the predictive value of genetic markers of growth response.

Method: We used pre-pubertal GHD children (peak GH <10 μg/l) from the PREDICT LTFU study (n=113) and PREDICT validation (VAL) study (NCT01419249, n=293). Single nucleotide polymorphisms (SNP) previously identified to be associated with first year growth response to GH (n=22) were genotyped. Random forest classification (RFC), a prediction method based on decision trees that is not sensitive to variable inter-dependency, was undertaken to identify variables associated with growth response (change in height (cm)) using the baseline clinical variables of gender, age, GH dose, distance to target height SDS (DTH) and mid-parental height SDS (MPH). Accuracy ((true positives+true negatives)/total population) of the RFC models was assessed and a variable importance score (VIS) calculated by permutation. GH peak was used to stratify GHD patients into severe (≤4 μg/l) and mild (>4 & <10 μg/l).

Results: Growth response in GHD severity-stratified sub-populations can be predicted by random forest classification with high levels of accuracy; in mild GHD an accuracy of 74.9% (P<3.0×10−17), in severe GHD an accuracy of 74.0% (P<7.3×10−15). Only baseline clinical variables were important in mild GHD with only GH dose and MPH (ranked by VIS) contributing to prediction. However in severe GHD VIS ranked important variables as followed: DTH, SNP rs1024531 (GRB10), age, SNP rs7101 (FOS), MPH, SNP rs3213221 (IGF2), and GH dose.

Conclusion: Growth response to GH therapy can be predicted by RFC using baseline clinical parameters alone in mild GHD. Three genetic markers (SNPs) can be used to improve growth response prediction in severe GHD.

Conflict of interest: Dr Adam Stevens has received honoraria as an investigator from Merck Serono.

Funding: The PREDICT study was supported by Merck Serono S.A – Geneva, Switzerland.

Article tools

My recent searches

No recent searches.