ESPE Abstracts

Background: A growth prediction model would not only allow patients with growth disorders the opportunity to see the expected effect of their recombinant human growth hormone (r hGH) treatment, but also support healthcare professionals to individualise treatment to optimise growth outcomes.

Aim: To develop a growth prediction model in children with growth disorders.

Patients and Methods: Height and clinical/background characteristics from children receiving r-hGH during the easypod™ connect observational study (ECOS)¹ were used to develop the model. For these children, height measurements from the easypod™ connect ecosystem were also included. Inclusion criteria were growth hormone deficiency (GHD), small for gestational age (SGA) or Turner syndrome (TS), age 2-18 years, age at treatment start 2-15 years, height standard deviation score (HSDS) at start <-1, ≥1 height measurement between 0.5-3.5 years’ treatment and treatment-naïve. A linear mixed-effects baseline model was applied, with HSDS between 0.5-3.5 years (y) as the dependent variable, and the main effects and interaction between the cubic polynomial of treatment duration and HSDS and age at start as fixed effects. Additional factors known at treatment start (e.g. sex) were then included in the model, which was validated using height measurements from patients from easypod™ connect who were not included in the ECOS study.

Results: In total, 1,212 children (885 with GHD, 243 with SGA, 84 with TS) with 7,485 measurements were available for model development. HSDS and age at start were highly significant in the baseline model (P < 0.001). The proportion of variance explained was 71%. Other significant factors (P < 0.01) were sex, corrected target HSDS, r-hGH dose (mg/kg/week) at start, disorder (TS vs SGA and GHD), puberty status (Tanner stage 1 vs >1) at start, birthweight SDS, and region (Western Europe, Asia/Oceania combined, Eastern Europe, Latin America and Caribbean, North America), but the proportion of explained variance only slightly improved to 72%. In the baseline model, predicted minus actual HSDS were 0.3 SD (1.9cm) at 1y, 0.5 SD (3.0cm) at 2y and 0.6 SD at 3y (3.9cm) after treatment start. Comparative figures were 0.3 SD (1.2cm) at 1y, 0.4 SD (3.2cm) at 2y and 0.5 SD at 3y (3.2cm) for 725 non-ECOS patients from easypod™ connect.

Conclusion: We developed a simple model, with age and height at treatment start, that accurately predicts future growth at any time between 0.5-3y after treatment start and at any starting age between 2-15y.

Reference: ¹Koledova E, et al. Endocr Connect 2018;7(8): 914-923.