Background and Aims

The paediatric obesity epidemic has led to renewed interest in the development and validation of relatively simple methods for defining or diagnosing obesity which are suitable for clinical and epidemiological purposes (Reilly Int J Obes 2006;30: 595-597). Systematic reviews-which reviewed data up to the end of 2001(e.g. SIGN 69, www.sign.ac.uk)-concluded that the evidence base from diagnostic studies (including a previous collaboration between the applicant and ALSPAC, Reilly et al Int J Obes 2000; 24: 1623-1627) supports the use of a high BMI percentile or SD score as a means of diagnosing or defining obesity in children and adolescents. Defining obesity as a high BMI for age has high specificity and moderate sensitivity when a high fat mass is used as the reference definition of obesity. In addition, children with high fat mass are at significantly increased risk of a number of morbid conditions, so the definition is not just arbitrary but has biological/clinical meaning.

More recently, there has been increasing evidence that a high waist circumference for age may be a promising alternative definition of obesity for paediatric use (Reilly Int J Obes 2006; 30: 595-597) . At present however the diagnostic accuracy of a high waist circumference in children and adolescents is unclear- the optimal cut-off point in the waist circumference distribution is unknown , ie the cut-off which best defines children with high fat mass and those children at greatest risk of co-morbid conditions. We also lack evidence from paediatric studies which have made direct comparisons of the diagnostic accuracy of the BMI and waist circumference as all studies published to date have reported on one or the other index but not both in the same sample.

The aims of the present study are therefore to

a. Estimate the diagnostic accuracy (e.g. sensitivity, specificity, predictive values) of a high waist circumference for a high fat mass index (fat mass measured by DEXA adjusted for height; Wells & Cole 2002 Int J Obes 26: 947-952).

b. Use ROC analysis to identify the optimum cut-off point in the waist circumference SD distribution which best identifies children with high fat mass index.

c Make a direct comparison between the diagnostic accuracy of BMI and waist by estimating sensitivity, specificity and predictive values for the currently recommended definitions based on BMI (such as BMI above the 91st and 95th percentiles) and the optimum definition derived from the present study, aim b above.

Methods