Allometric multi-scaling of weight-for-height relation in children and adolescents: Revisiting the theoretical basis of body mass index of thinness and obesity assessment

Abstract

The body mass index (BMI), defined as weight in kilograms divided by height in meters squared, has been widely used to assess thinness and obesity in all age groups, including children and adolescents. However, the validity and utility of BMI as a reliable measure of nutritional health have been questioned. This study discusses the mathematical conditions that support the validity of BMI based on population statistics. Here, we propose a condition defined as allometric uni-scaling to ensure the validity of BMI as an objective height-adjusted measure. Any given centile curve, including the median curve, in a weight-for-height distribution should be approximated using power-law functions with the same scaling exponent. In contrast, when the scaling exponent varies depending on the position of the centile curve, it is called allometric multi-scaling. By introducing a method for testing these scaling properties using quantile regression, we analyzed a large-scale Japanese database that included 7,863,520 children aged 5-17 years. We demonstrated the remarkable multi-scaling properties at ages 5-13 years for males and 5-11 years for females, and the convergence to uni-scaling with a scaling exponent close to 2 as they approached 17 years of age for both sexes. We confirmed that conventional BMI is appropriate as an objective height-adjusted mass measure at least 17 years of age, close to adulthood, for both males and females. However, the validity of BMI could not be confirmed in younger age groups. Our findings indicate that the growth of children’s weight-for-height relation is much more complex than previously assumed. Therefore, a single BMI-type formula cannot be used to assess thinness and obesity in children and adolescents.

Age- and height-dependent bias of underweight and overweight assessment standards for children and adolescents

“…To align our analysis with the targeted quantile ( q ), we set τ equal to q /100 while applying the q -th quantile regression. A simple search algorithm ( 80 ) was employed for parameter identification (see Ogata et al ( 68 ) for details). See Figure 1 for an illustration of quantile regression used to analyze weight-for-height distribution.…”

“…Moreover, there is skepticism about the direct association between increased BMI and heightened cardiovascular and mortality risks ( 67 ). In this context, we have recently posed a fundamental question about the allometric uni-scaling relationship that typically links weight and height in children and adolescents ( 68 ). Specifically, we analyzed a large-scale Japanese database encompassing 7,863,520 children aged 5–17 using a newly introduced method to test scaling properties through quantile regression ( 68 ).…”

“…In this context, we have recently posed a fundamental question about the allometric uni-scaling relationship that typically links weight and height in children and adolescents ( 68 ). Specifically, we analyzed a large-scale Japanese database encompassing 7,863,520 children aged 5–17 using a newly introduced method to test scaling properties through quantile regression ( 68 ). Our investigation revealed remarkable multi-scaling properties in males aged 5–13 and females aged 5–11, converging to uni-scaling as they approached 17 years for both sexes, exhibiting a scaling exponent close to 2.…”