The concept of optimum BMI can be applied to populations. For countries such as the United States, where undernutrition is not as common as in developing countries,3 a BMI-distribution median of around 21 kg/m2 may be optimal (WHO, 2000). Population weight goals for obesity prevention in adults can also be stated in terms of decreasing the proportion that exceed the threshold of 30 kg/m2, although this goal includes both preventing new cases of obesity and reducing weight among those already over the threshold.
The same principles are appropriate for assessing the population of children in the United States in pursuit of the committee’s primary objective: to stop, and eventually reverse, current trends toward higher BMI levels. Also, as discussed in Chapter 2, there are particular concerns about the population of obese children becoming heavier. Achieving this objective would have the effects of reducing the mean BMI as well as decreasing the proportion of children and youth in the population that exceeds the threshold definition of obesity.
Available research does not currently allow the committee to define an optimum BMI for children and youth. It suggests, however, that future research toward this aim should be focused on defining the associations between BMI and objective measures of concurrent and future growth and between BMI and physiological and psychological morbidity, mortality, and health (Robinson, 1993; Robinson and Killen, 2001).
Analogous to the current practice for adults, the committee recommends the use of BMI for assessing individual and population changes in children and youth over time and in response to interventions. Population weight goals for childhood obesity prevention should be stated in terms of changes in the mean BMI and in the shape of the entire BMI distribution. Alternatively, goals can be stated in terms of decreasing the proportion of children or youth who exceed particular thresholds—e.g., 75th, 85th, 90th, 95th, or 97th percentiles of BMI for age and gender on the CDC BMI charts. In the absence of an appropriate evidence base, however, threshold goals are necessarily somewhat arbitrary and sacrifice substantial information about the rest of the distribution as well as substantial statistical power to detect differences between groups and over time (Robinson and Killen, 2001).