Aim: Several methods have been developed to predict the outcome of growth hormone (GH) therapy in children with growth hormone deficiency (GHD).
Methods: Over 50 factors for each of 92 prepubertal patients with GHD (26 patients with total and 45 patients with partial GHD, 21 patients with neurosecretory dysfunction) were collected and included in multiple regression analyses and other nonlinear models to predict height velocity (HV) (cm/yr) in the first year of treatment. Afterwards the model was validated by two other cohorts of patients from other universities, which followed the same treatment regime as our clinic.
Results: Twelve parameters had a significant correlation to HV (p<0.05) and a coefficient of determination >20%. Two parameters (In BA, In GHmax) showed a coefficient of determination >60% for children with GHD in multiple regression analysis. The validation of the mathematical model against another data set showed different results. GH was measured by the same method, but BA was scored at the first clinic as in our clinic by only one doctor and in the other university by several radiologists. The accuracy of prediction in the first clinic was significantly higher than in the other university.
Conclusion: This model demonstrates that exact scoring of BA and precise measurement of GHmax in the stimulation test is necessary, and if carried out carefully leads to useful prediction values for determining height velocity.