Intended for healthcare professionals


Body mass index standards for children

BMJ 1998; 317 doi: (Published 21 November 1998) Cite this as: BMJ 1998;317:1401

Are useful for clinicians but not yet for epidemiologists

  1. Andrew M Prentice, Head of energy metabolism
  1. MRC Dunn Clinical Nutrition Centre, Cambridge CB2 2DH

    The body mass index (weight (kg)/(height (m)2)) is widely accepted as providing a convenient measure of a person's fatness. It gives an index that is broadly independent of height and equally applicable to men and women. A few individuals who are exceptionally muscular may be misclassified as overweight or obese, but otherwise the body mass index provides a rather robust index which has proved exceptionally useful for large scale epidemiological work. Its use is rapidly spreading into adult clinical medicine, where several charts and nomograms are available. Similar charts now exist for children, but their use is less straightforward.

    For adults a pragmatic classification system exists based on associations between body mass index and all cause mortality.1 The recently redefined body mass index categories are: underweight <18.5; ideal 18.5-24.9; pre-obese 25.0-29.9; obese class I 30.0-34.9; obese class II 35.0-39.9; and obese class III >40 kg/m2. These fixed classifications are not appropriate for children, in whom the 50th centile for body mass index shows profound changes from birth through to early adulthood (see figure). This has generated a lively debate on how best to measure the fatness of children, especially in relation to classifying obesity. The debate can usefully be separated into clinical and survey issues.


    Body mass index centiles for boys (those for girls are very similar)

    The clinical problem finds a ready solution in the 1990 British charts for body mass index spanning birth to 20 years, recently published by the Child Growth Foundation.2 These charts include nine centile curves based on divisions of two thirds of a standard deviation, and thus ranging from the 0.4th to 99.6th centile. Each card contains charts for girls and boys, although the gender differences are remarkably small.

    The new charts were created using data combined from several UK surveys conducted in 1980-90 and covering almost 15 000 children.3 The centiles were fitted using Cole's LMS method, which adjusts the body mass index distribution for differing degrees of skewness at different ages.4 The charts can therefore be confidently recommended for clinical application in the United Kingdom, and they will probably be widely adopted. They are less affected by differences in the timing of puberty than simple height and weight charts, but care must be taken not to confuse heavy musculature with obesity in a minority of children.

    So far, so good. Where the debate becomes complex is in relation to how to apply body mass index when trying to make secular or cross cultural comparisons between large epidemiological datasets for children. The rapidly developing epidemic of adult obesity in most affluent nations, and the consequent need to be able to monitor changing levels of fatness in children, gives urgency to the debate. This prompted the World Health Organisation International Obesity Task Force (IOTF) to work with the European Childhood Obesity Group (ECOG) to formulate a solution. The group, under the chairmanship of Professor Bill Dietz from Boston, is still considering its recommendations.5 At first sight the problem seems readily soluble. We simply need to choose a reference population and define some appropriate centile cut offs. However, this raises several problems.

    First is the inevitable political tussle about which population might best be used as the reference. In this respect the United States, whose National Center for Health Statistics' growth curves have dominated paediatrics for many years, is surely out of the running since it leads the world in obesity. The possibility of generating a globally pooled database has been examined. This reveals that the morphometric differences in the timing of height and weight growth among some populations (especially among Asiatics and South Americans) generate very different patterns of body mass index which could create serious anomalies in classification of obesity at certain ages. Aggregated centiles which exclude some of the atypical growth curves will shortly be published.

    Second, and most importantly for epidemiologists, is the problem of monitoring change over time. If growth curves are regularly updated to account for secular changes in nutrition then 10% of the population will always be above the 90th centile. To overcome this it is necessary to identify a reference dataset collected at a specified time. Ideally this would be several decades ago, before the serious emergence of obesity, but this is not essential. The Child Growth Foundation charts could be used to make both retrospective and prospective comparisons of secular change pegged to the British 1990 measurements. Alternatively the forthcoming IOTF charts could be chosen, but whatever the choice it is important to establish a reference fixed in time.

    Third is the problem of identifying health based cut offs for categorising obesity and underweight. In adults these are based, albeit crudely, on known risk ratios for different levels of body mass index. No such data exist for children. A possible solution to this dilemma has been suggested by the IOTF/ECOG working group.5 It involves identifying the centiles corresponding to the adult cut offs of 20, 25, and 30 kg/m2 and extrapolating back to childhood. The validity of this approach is currently being explored in a variety of datasets from around the world. The solution to these issues is by no means trivial, and the IOTF/ECOG recommendations are awaited with interest.


    View Abstract