ESPE Abstracts

Background/Aim: Parents wanting to predict adult height (AH) often resort to the old practice of doubling a boy’s height at age 2 years or a girl’s height at 18 months. We coined this the ’Grandma prediction’ (GMP1). It provides predictions with mean absolute error (MAE) 5.9 and 5.2, standard deviations of residuals (sdRES) 4.1 and 3.6, and Pearson correlation r=0.68 and 0.67 for boys and girls, respectively. The recent sophisticated machine learning methods are only modestly superior. Here, we aimed to optimize ‘Grandma prediction’, while keeping its simplicity.

Material & Methods: GrowUp1974 (2339 subjects, 1174 boys) and GrowUp1990 (1890 subjects, 964 boys) Gothenburg cohorts were used. Heights were imputed on a regular age grid (0-17.5 years, with 2 months step) using the Schumaker shape-preserving spline interpolation. We defined a new variable HAHA calculated as Half-Adult-Height Age per each subject. Given country-specific growth charts (Sweden, Germany, Belgium, Netherlands), we call “mean HAHA” (mHAHA) the age when a “zero-SDS” girl(boy) reaches her(his) Half-Adult-Height. We modify ‘Grandma prediction’ by optimizing the relevant parameters using GrowUp1974 and validate the predictions using GrowUp1990. Optimized GMP2 provides AH predictions by doubling the height at mHAHA of the relevant country, while GMP3 doubles the height at the age of 24 (18) months for boys(girls) as in GMP1, however corrects the prediction for target height.

Results: On average Swedish boys (girls) of either Gothenburg cohort reached their HAHA at age 27.3 ± 3.1 (20.6 ± 2.5) months. The mHAHA estimated from the Swedish growth charts were close to these values. Compared to ‘Grandma prediction’ GMP2 improves to MAE = 3.9 (3.9), sdRES = 3.0 (2.9), and r = 0.71 (0.68) for boys (girls). GMP2 is deemed appropriate for parents’ use after adjustment by mHAHA, estimated from the available country-specific growth charts. For example, we estimate the mHAHA to be 2.1 (1.6) years in Germany, 2.15 (1.65) years in Belgium, and 2.2 (1.8) years for Dutch boys (girls) in the Netherlands. The optimized GMP3 provides MAE = 3.9 (3.9), sdRES = 3.0 (3.0), and r = 0.70 (0.68) for boys (girls).

Conclusion: HAHA might become a new milestone in childhood at around 2 years of age, when adult height can be predicted by the GMP method. GMP2 proves useful for parents, while GMP3, corrected for target height, might be useful for pediatricians. Thus, the X-Rays-independent GMP2 and GMP3 are easy and practical methods for AH prediction.