Defining child obesity: a new step forward

Surveys of childhood obesity have been and continue to be immensely valuable, not least to identify the rapidly rising prevalence rates in lower income countries around the world over the last decade, and most recently to identify rising obesity levels among children during the COVID-19 epidemic. 

But researchers conducting these surveys have been faced with a difficult problem: whereas a body mass index (BMI) above 30kg/m² is generally accepted as a useful threshold for surveys among adults, for children BMI changes continually as they grow, so a BMI threshold to define obesity needs to vary month-by-month.

Over the last thirty years, national organisations have developed their own answers to this problem, and two international agencies have each developed a set of BMI thresholds (sometimes called cut-offs) for defining obesity at every age through to adulthood: the World Health Organization (WHO) and the International Obesity TaskForce (IOTF).

The problem for survey analysts is whether to use the IOTF or WHO versions, because they each give a different answer. The IOTF approach took half a dozen major surveys of children and tracked the proportion of children above BMI 30 at 18 years and tracked this as a constant proportion back thought to age 2 years, noting at each age (and for each gender) the BMI which gave the same proportion of children. The WHO approach took a definition used from a sample of healthy children at age 5 and defined obesity as the 2nd standard deviation of BMI above the median, and linked this to BMI 30 at age 20 years, and smoothed the connecting curve through the intervening years (again for each gender).

The two approaches give different results, and in general using the IOTF thresholds to define obesity in children will give a lower prevalence than using the WHO thresholds, creating difficulties when wanting to combine or contrast the results from surveys that have used different methods. For the last two decades, researchers have been urged to analyse their survey data using both methods and report both sets of results, to help avoid this problem.

Now a new paper* from the original author of the IOTF method, Professor Tim Cole, provides an algorithm that has the potential to take the prevalence values reported using one set of thresholds and convert them to prevalence levels using the other, in a reversible process. This could be of significant benefit to researchers modelling obesity development around the world, as it allows many more prevalence results to be combined to make a more comprehensive and robust estimate of trends, and strengthens the quality of predictive estimates.

Furthermore, it may have significant value in policy development and obesity services management. The two most influential research groups analysing obesity trends world-wide are (i) the Institute for Health Metrics, which produces Global Burden of Disease estimates and which bases their child obesity estimates on the IOTF approach, and (ii) the NCD Risk Factor Collaboration, which uses the WHO approach, and whose publications are widely used by WHO. The two research groups provide different assessments of the prevalence of childhood obesity globally and how they may change over time.

With the new algorithm there is an opportunity for modellers to start synthesising the resukts of surveys using either methods for assessing child obesity prevalence to create one that uses surveys from a wider range of sources. This is likely to give greater accuracy for making predictions, which helps policy makers to justify the development and implementation of interventions designed to prevent increases in obesity. For health service providers, more accurate assessment of current and future needs will improve the planning and management of child and adolescent services.

By Dr Tim Lobstein

* Exploring an algorithm to harmonize International Obesity Task Force and World Health Organization child overweight and obesity prevalence rates. Cole TJ, Lobstein T. Pediatr Obes. 2022 Feb 22:e12905.  doi: 10.1111/ijpo.12905. (open access paper)