LMS growth charts and sitar growth curves – old dogs and new tricks

Tim Cole

UCL Great Ormond Street Institute of Child Health, London, United Kingdom of Great Britain and Northern Ireland

The distinguished auxologist James Tanner greatly advanced the field of human growth and development. He popularised the construction and use of growth charts – the 1966 Tanner-Whitehouse standards set a high bar for what came later. However, his way of constructing skew weight centiles was both subjective and time-consuming, until the LMS method (Cole and Green, 1992) simplified the process. Later GAMLSS (generalised additive models for location, scale and shape) (Rigby and Stasinopoulos, 2005) provided a flexible toolbox for fitting centiles to data. Growth charts ideally need a reference sample of 7000 or more (Cole, 2021) – a large number to find when working with syndromic conditions, and infeasible with gene- specific charts for monogenic neurodevelopment disorders where there may be fewer than 100 patients worldwide. The talk will describe a modification to the LMS method for very small samples. Called the LMSz method, it uses an existing baseline growth reference to transform the data to z-scores, then models them using GAMLSS assuming a Normal distribution with linear age trends in the mean and standard deviation, and then back-transforms the centiles. The LMSz method will be applied to example data for several different genes, in work with Dr Karen Low (Bristol University). Tanner also experimented with growth charts modified to handle the chaos of puberty, where growth tempo is very variable between individuals. Longitudinal data are essential here to obtain a full picture, and the SITAR method of Cole et al (2010) summarises individual growth curves in terms of final height, peak height velocity and age at peak velocity. However, just as with growth charts, weight is more complicated than height as it fails to plateau post- puberty – it keeps on rising. The talk will describe an extension to the SITAR method which models the post-growth period of the weight curve, in work with Dr Ahmed Elhakeem (Bristol University) with examples from large cohort studies.