Validation of the KIGS growth prediction models in an independent cohort of a tertiary growth centre

Abstract

Background: Prediction models derived from growth data of patients documented in the Kabi International Growth Study (KIGS) need to be validated in independent cohorts to confirm applicability in clinical practice. Objectives: Retrospective comparison of predicted and observed first year growth response in children treated with recombinant human growth hormone of different brands. Patients and Methods: 597 children treated in the Hormone Centre for Children and Adolescents, University Hospital of Schleswig-Holstein, Kiel, Germany, were considered. 230 patients met the requirements for use of prediction models. The model for Growth Hormone Deficiency (GHD) was validated in 141 patients (classical idiopathic growth hormone deficiency (IGHD): n = 39, neurosecretory dysfunction (NSD): n = 102), that for Ullrich-Turner-Syndrome (UTS) in 46 patients and that for Small for Gestational Age (SGA) in 43 patients. NSD and IGHD as well as the six various brands were also analysed separately. Results: Low mean absolute Studentized Residuals (SRs) indicate that differences between predicted and observed growth responses are small for all diagnoses (IGHD: 0.96 ± 0.75, NSD: 0.64 ± 0.61, UTS: 0.84 ± 0.60, SGA: 0.87 ± 0.73) and for the different brands (A: 0.93 ± 0.77, B: 0.68 ± 0.48, C: 0.92 ± 0.82, D: 0.57 ± 0.44, E: 0.67 ± 0.69, F: 0.86 ± 0.61). Positive correlations between predicted and observed height velocity reflect their causal connection (IGHD: r = 0.72, NSD: r = 0.36, UTS: r = 0.43, SGA: r = 0.43). Conclusion: KIGS prediction models for GHD, UTS and SGA are applicable in an independent cohort. Prediction works regardless of the applied brand of recombinant human growth hormone.