Prediction of Adult Height Based on Automated Determination of Bone Age

Abstract

Context: Adult height prediction is a common procedure in pediatric endocrinology, but it is associated with a considerable variability and bias from the bone age rating. Objective: A new method for adult height prediction is presented, based on automated bone age determination. Method: The method predicts the fraction of height left to grow from age and BoneXpert bone age. This is refined by drawing the prediction toward the population mean, or alternatively toward the height predicted from the parents’ heights. Boys’ body mass index and girls’ height at menarche can be included optionally as predictors. Participants: A total of 231 normal children from the First Zurich Longitudinal Study (1ZLS) were followed from age 5 until cessation of growth with annual x-rays of the left hand. A total of 198 normal children from the Third Zurich Longitudinal Study were used for validation. Results: The root mean square error of adult height prediction (Tanner-Whitehouse 3 method in parentheses considered as standard for accuracy) on the 1ZLS was 3.3 cm (3.5 cm) for boys aged 10–15 yr and 2.7 cm (3.1 cm; P < 0.005 for difference to Tanner-Whitehouse 3) for girls aged 8–13 yr. High body mass index before puberty negatively affected adult height of boys, independent of bone age. Conclusions: With the new method, adult height prediction has become objective because the dependence on manual bone age rating is eliminated. The method is well-suited to analyze large studies and provide a consistent body of evidence regarding the relation between maturation, body mass, and growth across populations, conditions, and ethnicities.