Must and Anderson,1 in this issue of the journal, have described the basis of body mass index (BMI) for age for the assessment of weight status in children and adolescents. The aims of this short article are to summarise the evidence base on the diagnostic accuracy of BMI, and to expand on what we achieve when defining paediatric obesity on the basis of a high BMI for age. A high BMI for age is not 'obesity' (a high body fat content associated with increased morbidity), per se but a diagnostic test for obesity. Similarly, in adults a BMI>30 kg/m2 is not obesity, but a positive diagnostic test for obesity. It is also widely stated that 'BMI does not predict body fat content accurately'. This is true,2, 3, 4, 5 but should not be used to damn the diagnostic ability of BMI because it is not directly relevant to the issue of diagnosis of obesity. When diagnosing or defining obesity, we do not need to estimate body fat content precisely; rather we simply need to establish with confidence that any child defined as obese has a high body fat content relative to his/her peers. There are two main components to the issue of diagnostic accuracy: how well does a high BMI for age diagnose the fattest children in the population? Are children at high BMI for age at greater risk of morbidity than their peers at lower BMI for age