A prediction model could foresee adequate height response in children eligible for growth hormone treatment.

Aim: Prediction models may be useful in accurately identifying children who will benefit from growth hormone (GH) treatment. We aimed to validate the Gothenburg prediction model for this purpose.

Methods: The study included prepubertal children with GH deficiency who started treatment with GH during 2004-2016 at Queen Silvia Children's Hospital, Gothenburg, based on a first-year growth prediction of ≥0.7 SDS in height according to the Gothenburg prediction model on a GH dose of 33 μg/kg/day. Observed heights retrieved from medical charts were compared with predicted heights.

Results: The study included 121 patients (64% boys) with at least one year of data after the start of GH treatment. The median (range) GH dose was 30 (10-43) µg/kg/day for the first year and age at start of treatment was 5.32 (3-11.8) years. The model correctly excluded poor responders resulting in 119/121 (98%) patients having a height gain of at least ≥0.5 SDS in a year. However, the model underestimated relatively low predictions and overestimated relatively high predictions, with a residual standard deviation of 0.31 SDS.

Conclusion: By using a validated prediction model for GH in a clinical context, unnecessarily treating short children with an expected poor height outcome can be avoided.