Endocrine Abstracts

The pubertal growth spurt involves a rapid increase then decrease in height velocity (HV), where both the timing and intensity of the spurt vary considerably between individuals. The analysis of such growth curve data is an ongoing statistical problem. Current approaches involve fitting say the Preece–Baines (PB) curve to each individual’s data, and summarising the curves by averaging the five fitted PB parameters across individuals. But this has two drawbacks: fitting lots of separate curves, and needing as many as five parameters per individual.

The talk will present an alternative model of pubertal height growth, due to Beath and called SITAR, where just one growth curve is fitted that applies to all individuals. Growth curves for Individuals are matched to the average curve by shifting their curve up-down (representing differences in pre-pubertal size) and left-right (for differences in tempo or age at peak HV), and the age scale is also stretched or squashed (indicating how fast time passes in the individual, i.e. velocity). These three parameters per individual are estimated as random effects at the same time as fitting the average curve, and the goodness of fit of the model matches that of individually-fitted PB curves. The outcome is a cubic spline growth curve that applies to all subjects, plus triplets of parameters per individual (size, tempo and velocity) that summarise their departures from the mean curve.

Two datasets were used to demonstrate the method: N=3245 boys from Christ’s Hospital (CH) School recruited 1939–68 and measured twice a term from 9 to 19 years (n=129 508), and N=92 Turner syndrome (TS) girls from the UK Turner Study measured 6-monthly from 9 to 16 years (n=1102).

The growth curve model explained 99% of the variance in both datasets (RSD 6–7 mm). In CH, growth tempo (but not size or velocity) was strongly associated with IGF1 measured 50 years later (P=0.0004, N=1014). The TS girls randomised to receive oxandrolone from 9 years did not differ from the placebo group in terms of size or tempo, but their velocity was highly significantly increased (P<10⁻⁸).

The SITAR growth curve model is a major step forward in the analysis of pubertal growth.