Recently published standards for body mass index (BMI) based on population studies of height and weight in healthy British children allow an easy but indirect assessment of adiposity in healthy children. However, assessment of adiposity based on standards derived from reference populations may not be appropriate for use in subjects with disease states associated with abnormalities of growth and body composition. This hypothesis was tested by comparison between BMI standard deviation scores (SDS) and more direct measures of body fat derived from dual-energy X-ray absorptiometry (DEXA) and skinfold thickness in groups of children, receiving growth hormone, with inflammatory bowel disease, previously treated for malignancy, and healthy controls. Excess adiposity was defined as a body fat greater than the 85th percentile and was compared to a BMI SDS of +1.0. Overall the sensitivity and specificity for a BMI SDS of +1.0 to correctly identify individuals as having excess adiposity was 66% and 94%, respectively, when body fat was measured by DEXA, and 50% and 100% when estimated from skinfold measurements, respectively. There were no significant differences in these statistics whether applied to the individual disease groups or to healthy controls. These findings suggest that BMI under-predicts the prevalence of excess adiposity in children with disease states but surprisingly to no greater degree than that seen in healthy subjects.