Title of article:
Identification of the obese child: adequacy of the body mass index for clinical practice and epidemiology
Authors: Reilly JJ, Dorosty AR, Emmett PM, et al
Journal: Int J Obes Relat Metab Disord, Dec 2000;24(12):1623-7
Abstract
OBJECTIVE: To assess the ability of simple definitions of BMI to successfully screen for children with high body fatness. DESIGN: We determined the sensitivity and specificity of the body mass index (BMI) by testing its ability to correctly identify children with high body fat percentage. Receiver operator characteristic (ROC) analyses were carried out using the top 5% of body fat percentage to define children as obese (true positives). SUBJECTS: Representative sample of 4175 7 y-old (88-92 month-old) children (2120 boys; 2055 girls) participating in the Avon Longitudinal Study of Pregnancy and Childhood (ALSPAC). RESULTS: The current obesity definition based on BMI (95th centile) had moderately high sensitivity (88%) and high specificity (94%). Sensitivity and specificity did not differ significantly between boys and girls. The ROC analysis showed that lower cut-offs applied to the BMI improved sensitivity with no marked loss of specificity: the optimum combination of sensitivity (92%) and specificity (92%) was at a BMI cut-off equivalent to the 92nd centile. Sensitivity of BMI using the new International Obesity Task Force (IOTF) cut-off for obesity was much lower, and differed significantly (P < 0.001) between boys (46%) and girls (72%). CONCLUSIONS: Screening for childhood obesity using the BMI is specific, and can have moderately high sensitivity if an appropriate cut-off is chosen. New recommendations based on the IOTF approach to defining childhood obesity are associated with lower sensitivity, and sensitivity differs between boys and girls.
Comments and Key points
This seems like a good article, but upon closer inspection it was not helpful. It's key limitation is that it only examined 7-year old children, not other ages.
It defined "true obesity" as children with body fat percentage (measured by impedence) above the 95th percentile of their sample. This definition may or may not be OK, but it is arbitrary.
This study compared BMI thresholds to see which gave the most "sensitivity" for detecting "true obesity" correctly. They compared 2 types of thresholds, which I call type A and type B.
For type A, they quoted a BMI (for 7-year olds) of 19.3 for boys and 19.2 for girls as obesity thresholds, based on the IOTF definition1. These values are quite close to, but slightly lower than- the CDC growth chart BMI values. Unfortunately, the reference they give for IOTF definition1, does not seem to match these values ( 19.3 and 19.2). Instead, that reference1 says 20.63 and 20.51! So I'm just totally confused because I cannot verify which criteria they actually tested for type A. Anyway, they concluded that type A criteria weren't very good.
For type B, they referred to a UK dataset2,3 of body mass index, and used the 95th percentile BMI threshold value as a BMI obesity threshold. Unfortunately, their article never bothered to actually mention the actual values they used, forcing the reader to have access to the original dataset.
Finally, the study concluded that the best sensitivity and specificity BMI threshold occurred at the 92nd percentile fo the UK BMI dataset. Unfortunately, their article never actually mentioned the actual BMI values of these 92nd percentiles. Incredibly frustrating.
So the reader of the article learns methods but no usable facts.
References
- Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Body mass index in children worldwide: cutoff points for overweight and obesity. Br Med J 2000; 320:1-6.
- Barlow SE, Dietz WH, Obesity evaluation and treatment: expert committee recommendations. Pediatrics 1998; 102:c29
- Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990; Arch Dis Child 1995; 73: 25-29
Review & comments by Steven B. Halls, MD, Last edited 23-June, 2003, Copyright
Back to scientific BMI articles.
