Identifying very fat and very thin children: test of criterion standards for screening test

BMJ 1999; 319 doi: (Published 23 October 1999) Cite this as: BMJ 1999;319:1103
  1. J Mulligan, data manager (jm1{at},
  2. L D Voss, senior research fellow
  1. University Child Health, Southampton General Hospital, Southampton SO16 6YD
  1. Correspondence to: J Mulligan

    Charts of body mass index (weight (kg)/(height (m)2)) for children have been recommended for clinical use in the United Kingdom.1 It is unclear how they are to be used, but the accompanying referral guidelines suggest that they have a place in community screening. Caution has been urged, however, when using body mass index for children of different heights.2 Data from the Wessex growth study show that 10% of short normal children—compared with 27% of controls—had a body mass index ≥75th centile at the age of 9 years (unpublished data). To determine the reason for this apparent bias, we examined the weight for height relations required for prepubertal children of different heights to have a high or low body mass index.

    Methods and results

    Weights and weight standard deviation scores were calculated for children of different heights to give body mass indices ≥99.6th centile and ≤0.4th centiles. The difference between height and weight centiles was then calculated. The table shows data for boys aged 2-9 years with heights on the 99.6th, 50th and 0.4th centiles. Results for girls were similar.

    Relation between height and weight in tall, average, and short prepubertal boys for a body mass index (wt/ht2) equal to the 99.6th centile and the 0.4th centile)

    View this table:

    To have a body mass index on the 99.6th centile the discrepancy between weight and height for a child of average height is around three centile bands, regardless of age. For children of other heights, however, the discrepancy varies with height and age. For example, the discrepancy for a tall child ranges from 1.67 centiles at age 2 years to less than half a band at 9, while short children require weight to be at least four centiles above height to have the same body mass index score. To have a body mass index ≤0.4th centile, however, the discrepancy between weight and height for a child of average height is less than three centiles, while tall and short children require larger or smaller discrepancies (table).


    Body mass index is well established as a measure of relative fatness in adults, but during childhood it changes substantially with age and must be assessed using age related curves such as those provided on the charts. Body mass index, however, correlates with height,3 short children having lower body mass indices, and thus lower body mass index standard deviation scores, than tall ones. Using body mass index to assess weight for height in the community may identify a disproportionate number of tall, apparently overweight, children, while short overweight children will be harder to detect.

    Other methods of identifying weight disorders have been proposed—for example, a discrepancy between weight and height of more than three centiles.3 Although there is a general acceptance that healthy children have similar height and weight centiles, our data show that overweight and underweight children cannot be identified by the same criterion. To have a body mass index ≥99.6th centile, the discrepancy between weight and height for a child of average height is three centiles, but to have a body mass index ≤0.4th centile, the discrepancy is less than three centiles. Furthermore, except for those of average height, the relation between height, weight, and body mass index varies considerably throughout childhood.

    There is as yet no agreed measure of obesity in children.1 The prevalence of obesity increases with age, and the centiles defining adult obesity are unlikely to yield a similar proportion of clinically obese children. Furthermore, targeting obese children is unlikely to identify those most at risk of becoming obese adults.4 Visceral fat distribution is likely to prove a better predictor of subsequent morbidity than absolute fat mass.5 The charts allow change in body mass index to be observed in an individual child, but this may be no more valuable than the longitudinal monitoring of height and weight from which body mass index is derived. Children with diverging height and weight centiles should perhaps be referred rather than waiting for body mass index to cross an arbitrary cut off point, especially one that has no proved clinical correlate.


    Contributors: JM devised the investigation, analysed the data, and contributed to the writing of the paper. LV was the coordinator and principal investigator of the Wessex Growth Study and contributed to the writing of the paper.


    • Funding The Wessex Growth Study is funded by a grant to the Wessex Medical Trust from Pharmacia Upjohn and by a grant from the NHS S and W research and development fund.

    • Competing interests None declared.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    View Abstract