ESPE Abstracts (2014) 82 FC7.4

A Decade of Clinical Experience in a Swedish University Centre Using Prediction Models to Optimize GH Treatment in Prepubertal Children

Jovanna Dahlgren


Department of Pediatrics, Göteborg Pediatric Growth Research Centre, The University of Gothenburg, Gothenburg, Sweden


Background: The individual growth response on a certain GH dose is an indirect measurement of tissue responsiveness to GH. Several models of predicting the first year growth response on GH treatment are published and a clinical trial has been performed based on these algorithms. However, no clinical unit has evaluated the practical use of these prediction models on GH treatment.

Objective and Hypotheses: To use prediction models in clinical practice to choose those short children that will benefit of GH treatment. Prediction models can be used as a complement to optimize GH treatment.

Method: The prediction model used is based on heightSDS and weightSDS at GH-start, early growth, mid-parental height SDS, 24-h GH profile, IGF1SDS. 122 short prepubertal children (60% boys) (height <−2 SDS) were investigated at Queen Silvia Children’s Hospital 2004–2012 with a 12/24-h profile and IGF1 levels. Exclusion criteria were chronic disease, malignancy, small-for-gestational-age and syndrome (n=20) or predicted height gain <0.7 SDS (n=39). Mean height SDS was −2.6 (range, −4.2 to −2 SDS).

Results: Mean predicted height gain the first 12 months on treatment was 0.85 SDS (range, 0.7–1.6). The observed height gain during these 12 months was 0.66 SDS (range, 0.5–1.3). This was studied with Bland–Altman plot. A difference between observed and predicted response of 0.20 SDS was found. Interestingly, the majority of children were systematically over-predicted 0.20–0.28 SDS, but still the individual diagnosis of very good or good responders before treatment was confirmed on treatment.

Conclusion: We found a good correlation between predicted and reported first year growth in short children treated with GH, although with a systematic overestimation of the individually predicted values. This is expected by previous validations (2×SDres of 0.37). Despite this overestimation, the model can be used as an integrated tool in decision-making at start of GH treatment.

Article tools

My recent searches

No recent searches.