A child’s growth spurt in height or weight can be summarised by just three numbers—size, timing, and intensity—but the statistical tool that extracts those numbers currently gives biased results when linked to later health outcomes like obesity or diabetes. The problem is that the SITAR method, which explains more than 99% of the variation in adolescent height, was designed to describe growth curves, not to test how those curves relate to adult disease. Researchers typically estimate the growth parameters in one step and then correlate them with outcomes in a separate step, ignoring the error structure and producing unreliable associations. SITAR also assumes growth plateaus after the spurt, so it fails to fit weight, body fat, lean mass, or bone mass, which continue changing into adulthood. This project will develop joint statistical models that estimate growth features and their effects on later health simultaneously, eliminating bias. It will also revise SITAR to handle post-peak growth and body composition curves. The new methods will be tested on repeated measurements from age 5 to 20 in four long-running UK, US, and Canadian cohort studies. Open-source R packages and guidance materials will let epidemiologists apply the tools across multiple studies. If successful, researchers will finally have an unbiased, accessible way to ask how early-life factors—prenatal stress, famine, recession—shape adolescent growth and, in turn, adult chronic disease risk.
View original technical description
Adolescence is characterised by pubertal growth spurts in height and weight, and rapid changes in body composition. These growth patterns can be influenced by early life factors (e.g., prenatal stress, famine, war, recession) and can have consequences for adult health (e.g., obesity, diabetes, cardiovascular disease, osteoporosis). The SITAR (Super Imposition by Translation and Rotation) method of growth curve analysis summarises individual growth patterns using three parameters (size, timing, and intensity) that are estimated as random effects, plus a cubic spline estimate of the average growth curve. SITAR was designed to simplify the analysis of adolescent height growth curves in individuals and explains >99% of the age-specific variance in height, making it an accurate and efficient method to summarize individual growth trajectories SITAR random effects can be analysed further in relation to earlier growth-altering exposures or later health outcomes, making the model highly relevant for translational medicine and life course epidemiology. However, these analyses are often performed in two stages–first estimate the random effects, then relate them to the exposure/outcome. This two-step approach leads to biased estimates of the association as it ignores the underlying random effect error structure. Moreover, SITAR assumes a plateau or constant growth at the end of the growth spurt which means it does not properly fit outcomes that continue to change, and at different rates, going into adulthood (e.g., weight, adiposity, lean mass, and bone mass). The aim of this project is to facilitate unbiased analysis on determinants and outcomes of height and body composition growth around adolescence and emerging adulthood and empower researchers with essential information and tools for the best-practice analysis of individual growth patterns and their correlates. The project will address the limitations described above by (1) tackling the outstanding methodological issues, (2) developing open-source software to implement the new insights and (3) creating resources to guide researchers with implementation and interpretation of the methods. Methodological developments include methods for joint models to simultaneously estimate SITAR growth features as exposures and the effects of those features on later outcomes. Complementary developments include revising SITAR to effectively quantify variability in post-peak growth rate and to accurately fit weight, adiposity, lean mass, and bone mass growth curves. The new methods will be tested using repeated growth measurements from age 5-20 years in four well-established prospective cohort studies in the UK, USA, and Canada, and using simulation studies. New R packages, a workshop, and interactive guidance material will enable statisticians and epidemiologists to apply the method relatively simply, including in multicohort study settings, and to interpret the results appropriately. This project makes an important novel contribution to modelling of longitudinal growth measures and will significantly increase our understanding of the determinants and outcomes of growth in adolescence and emerging adulthood. Together, the methodological developments will facilitate unbiased state-of-the-art analysis of individual growth curves and their correlates and make this accessible to researchers.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know