Variables and statistical analysis

Clinical end points included HSDS, BMI SDS, bone age per chronological age (BA/CA), insulin-like growth factor 1 (IGF-I) SDS and GH dose. HSDS and BMI SDS (z-scores) were calculated from actual height and weight using standard references provided by the US Centers for Disease Control and Prevention (24). IGF-I values were measured locally and transformed into IGF-I SDS based on age- and sex-related normative values of Brabant et al. (25). Reported means, including those referring to change over baseline values, were calculated as a mean of the patients within each subpopulation. Patients with missing birth weight and/or length information or with missing baseline values of key variables (height, BMI, BA, CA, gender and IGF-I) or any of the clinical end points were excluded. Only observed values were used in the analysis; missing values were not imputed. Participants with missing data for some, but not all, the follow-up visits were included for the years with available data.

All statistical tests were two-sided and P values ≤0.050 were considered statistically significant. Descriptive summaries were provided for baseline characteristics and clinical end points at each year of GH treatment and reported as mean (standard deviation (s.d.)) unless otherwise noted. Differences in ΔHSDS and ΔIGF-I SDS at each year of GH treatment were compared by analysis of covariance (ANCOVA), with diagnostic population, gender and baseline age as fixed effect covariates. Least-squares (LS) means for ΔHSDS and ΔIGF-I SDS at year 3 were estimated using the Tukey–Kramer test. All analyses were performed using SAS version 9.3 (SAS Institute).