Individualized Growth Hormone Treatment - Using Two Different Prediction Models in a Clinical Setting

Background: Diagnosing growth hormone deficiency (GHD) can be challenging; hence, prediction models on growth outcome from growth hormone (GH) treatment have shown to be useful. We aim to compare the accuracy of the more readily available KIGS (Pfizer International Growth Study) prediction model to the previously clinically validated Gothenburg model. Methods: Prepubertal children with GHD who started GH treatment at Queen Silvia Children’s Hospital between 2004 and 2016 were considered for the study. Exclusion criteria were short stature due to syndrome, chronic disease, oncology disease, or known bad adherence. Growth predictions were made according to the Gothenburg model and the KIGS model. Growth data from birth until one year after start of GH treatment were collected from medical charts. Predicted height and observed height were then compared. Results: A total of 123 children, 47 girls (38%) and 76 boys (62%) were included, with a mean age of 5.71 (±1.81 SD) years at start of GH treatment. The Pearson correlation of predicted first-year growth versus growth outcome were r = 0.990 for the Gothenburg model and r = 0.991 for the KIGS model. Studentized residuals were 0.10 ± 0.81 SD and 0.03 ± 0.96 SD, respectively, for the models. The comparison between the two models showed r = 0.995. Conclusion: The Gothenburg model and the KIGS model are equally accurate at predicting height outcome from GH treatment for our study cohort. We therefore promote the use of either model in clinical settings.