Modelling height in adolescence: A comparison of methods for estimating the age at peak height velocity

INTRODUCTION:  Accurate assessment of maturational status is an important feature in many analyses involving children and adolescents, and maturity is frequently used as an outcome, exposure, or to control for confounding. Currently there are a variety of different methods which can be used to assess maturity, including the age at peak height velocity (PHV). However, there are many different ways to estimate age at PHV.

METHODS:  Using a simulation study we investigated five different methods of modelling adolescent growth, and estimating the age at PHV. Methods included; central differencing, the Preece-Baines growth model, multi-level models with fractional polynomials, SITAR, and PACE. We explored the effect of measurement error, sample size, frequency, and measurement balance on the estimates of age at PHV and (where available) the confidence intervals and coverage around these estimates.

RESULTS:  The SITAR and Preece-Baines growth model were unbiased in the majority of experimental scenarios, however model convergence is problematic, especially for SITAR, when measurement error is large and measurements are infrequent. Nearly all methods demonstrated strong correlations (r>0.9) between true and estimated age at PHV.

CONCLUSIONS:  The use of parametric growth models to identify the age at PHV is challenging, and despite their quantitative appeal, substantial bias and variability can occur.