ESPE Abstracts (2018) 89 P-P3-212

Height Velocity and Height Gain in the First Year of GH Treatment: Predictive Factors of Good Statural Response in Small for Gestational Age (SGA) Patients

Régis Coutanta, Bruno Leheupb, Marc Nicolinoc, Jean-Pierre Sallesd, Evguenia Hacquese & Béatrice Villettee


aCHU, Angers, France; bCHRU – Université de Lorraine, Nancy, France; cHôpital Mère-Enfant – Groupement Hospitalier Est, Lyon, France; dHôpital des Enfants, CHU, Toulouse, France; eNovo Nordisk, Paris, France


Objective: Bang et al. showed that Δ height and growth velocity (GV) the first year of treatment could be predictive factors of statural response in SGA (n=54). Poor responders showed GV <1SDS (55%) and Δ height <0.5SDS (45%). Moreover, Ortego et al. seems to confirm the suitability interest of KIGS mathematical model in a retrospective SGA cohort (n=103) showing that the percentage of good responders in the first year varies between 46.6% (Δ height ≥0.5SDS) and 81.6% (GV≥1SDS). The authors questioned if these predictive factors for the final adult height (FAH)>−2 SDS could be applicable in the French SGA cohort.

Methods: Observational ongoing study of 291 SGA children treated with Norditropin® included 183 naïve patients. Naïve completers (n=51) were stratified to poor and good responders according to FAH ≤−2SDS or >−2SDS, respectively. Logistic regression model for prediction of FAH (≤−2SDS/>−2SDS) considered the Δ height or GV in the first year of treatment. The value of the area under the curve (AUC) defines the strength of the model to distinguish poor from good responders considering the value of explanatory variable (Δ height or GV): low predictive model if AUC<0.7; moderate predictive model if 0.7≥AUC<0.9; excellent predictive model if AUC=1.

Results: Δ height in the first year. The best prediction of good response (AUC=63.3%) was obtained by stratifying the variable in these classes: ≤0.5/>0.5 SDS (odds ratio (OR)=3, confidence interval (CI)=0.93; 9.70, P=0.0665). The concordance of observed and predictive FAH for good responders concerns 67% of patients (Table 1). The best prediction of good response (AUC=65.8%) was obtained by stratifying the variable in classes: ≤0.75/>0.75 SDS (OR=5.32, CI=1.35; 20.98, P=0.017). The concordance of observed and predictive FAH for good responders concerns 86.6% of patients (Table 2).

Table 1 The error rate of wrong categorisation of patients is 36%
Predictive values
Observed values<2 SDS>−2 SDS
<−2 SDS128
>−2 SDS1020
One patient with missing data
GV in the first year
Table 2 The error rate of wrong categorisation of patients is 30%
Predictive values
Observed values<2 SDS>−2 SDS
<−2 SDS128
>−2 SDS1020
One patient with missing data

Conclusion: The strength of this predictive model has not been confirmed perhaps due to small sample size. Nevertheless, there were some interesting observations for good responders. Further investigations are needed because this type of model might help in managing short stature patients.

Article tools

My recent searches

No recent searches.