Aims A variety of methods are available for defining under-nutrition (thinness/under-weight/under-fat) and over-nutrition (overweight/obesity/over-fat) in children and adolescents. The extent to which these definitions produce similar results is unclear. This study aimed to assess agreement between widely-used methods of assessing nutritional status, in rural South African children and adolescents, and to examine the benefit of body composition estimates rather than simple proxies for body fatness.

Methods In an age stratified random cross-sectional sample of 1519 boys and girls from rural South Africa, in three equal groups (school grades 1, 5 and 9 corresponding to ages 7, 11 and 15 years respectively), underweight, healthy weight, overweight and obesity were assessed using four methods: BMI-for-age using WHO (2007) reference data; Cole (2007) and Cole-International Obesity Task Force (IOTF) cut offs; weight-for-age (CDC-WHO 1977); % body fat estimated by impedance (Jebb 2004). Comparisons were made between the different methods using weighted kappa analyses and defined using Landis and Koch categories.

Results In both boys and girls, prevalence of unhealthy weight status (both under- and over-nutrition) was higher at all ages with body fatness measures than with simple anthropometric proxies for body fatness.

Prevalence of healthy weight status with body fatness measures in boys ranged from 36% at age 15 to 74% at age 7, and in girls from 63% at age 7 to 72% at age 11. With the other simple anthropometric proxies for body fatness prevalence of healthy weight status was much higher; 78-92% in boys and 69-91% in girls. Kappa agreement between fatness and weight-based measures was only fair or slight in the boys but fair to substantial in the girls using Landis and Koch categories.

Conclusions Methods for defining under- and over-nutrition should not be considered equivalent. Weight-based measures are easy to collect however they appear to provide highly conservative estimates of unhealthy weight status, possibly more conservative in boys than girls. Simple body composition measures may be more informative than anthropometry for nutritional surveillance of children and adolescents in low-middle-income countries.