The SITAR method of growth curve analysis for growth assessment in translational medicine and life course epidemiology

The pattern of growth in individual children, as represented by their growth curve of successive measurements (e.g. weight or height) plotted against age, may be informative about their circumstances at earlier ages, or conversely it may predict their risk of disease later in life, for example heart disease, stroke or type 2 diabetes. To investigate this it is necessary to summarise the shape of each individual's growth curve in a simple way, so that its shape can be related to their earlier or later experiences. The SITAR method of growth curve analysis is a novel approach that does just this - it summarises each individual growth curve in just three numbers, reflecting respectively their size (i.e. tall or short, heavy or light, as compared to others), their growth tempo (i.e. whether they are relatively mature or delayed in their timing of growth, e.g. age of menarche in girls), and velocity (how fast they are growing compared to others, e.g. in puberty). In addition SITAR estimates a smooth average growth curve against which each individual's growth can be compared. The surprising thing is that this model explains nearly all of the variability in growth between individuals, so that their size, tempo and velocity can be viewed as a complete summary of their growth pattern. Put another way, when adjusted for size, tempo and velocity all their growth curves are the same. It is then simple to relate the values of size, tempo and velocity in individuals to their earlier or later experiences, and to look for links between them.

The SITAR method of growth curve analysis summarises individual growth patterns of e.g height or weight in three parameters (size, tempo and velocity) that are estimated as subject random effects, plus a cubic spline estimate of the average growth curve. It explains over 95% of the age-specific variance in the outcome measure, and as such is an effective summary of individual growth patterns. The subject-specific random effects can then be related to earlier growth-affecting exposures or later health outcomes. Thus it is relevant in translational medicine and life course epidemiology. The aims of the project are (i) to further develop the SITAR methodology, (ii) to apply it to a series of existing growth studies, and (iii) to develop a SITAR software library in the R statistical language for its wider dissemination. Methodological developments will include exploring P-splines as an alternative to B-splines for the average curve, and deriving standard errors for the random effects. The application studies are collaborations involving a mix of observational and experimental data, covering height, weight and BMI in infancy, puberty and all childhood, and they include the British 1946, 1958, ALSPAC and Millennium birth cohorts. The R library will enable other statisticians and epidemiologists to apply the method relatively simply.

SITAR can also be applied to cross-sectionally based median curves of say height versus age, as used in growth charts. SITAR allows such median curves from different growth references to be superimposed, which is useful for merging charts and for understanding differences in the underlying reference populations.

Who will benefit from this research, and how?

o Academia
All researchers, national and international, interested in the analysis of longitudinal growth will find SITAR of interest, as it will allow them to summarise individual growth patterns in a simple way and then link the pattern to other aspects of the individual's health and lifestyle.

o Public Sector
National policy makers concerned with the high and rising levels of childhood obesity will want to explore the idea from SITAR that overweight children have an advanced maturational age, and look for lifestyle interventions that might counteract this accelerated age pattern.

o Business/Industry
The international pharmaceutical industry will value a statistical tool that simplifies and sharpens the assessment of growth-affecting treatments. SITAR's efficient use of data will mean that clinical trials involving growth outcomes need not be as large as before, and hence may reach conclusions more quickly and certainly more cheaply. This in turn should make cheaper the development of new growth-affecting drugs.

o Parents and children
Parents and children worldwide will benefit from improved growth assessment through the development of more effective evidence-based growth charts, particularly in puberty. Improvements in the assessment and treatment of obesity will impact on the quality of life of children now and on the adults they will become.