Interpreting Weight, Height, and Body Mass Index Percentiles in the US Centers for Disease Control and Prevention Growth Charts

To the Editor:

Hendrickson and Pitt¹ expressed concern that a child whose height and weight are both at the US Centers for Disease Control and Prevention (CDC)–defined 97th percentile would have a body mass index (BMI) above the 85th percentile of the CDC growth charts² rather than in the normal weight range. Although the authors suggested that this seemingly counterintuitive observation may be the result of different data sets in the growth charts or a weakness of the BMI formula, the association of BMI z score with to z scores for weight and height was addressed by Cole³ in 2002. I further examined the associations between these sex- and age-standardized z scores for weight, height, and BMI and compared the association of body fat with both BMI z score and a metric based on the weight percentile minus height percentile difference (WHD).

I used data from National Health and Nutrition Examination Survey 2011 to 2012 through 2017 to 2018⁴ for the analyses, which included 13 042 individuals aged 2 to 19 years with weight and height data. Body fat was calculated from dual-energy radiography absorptiometry for 6923 male individuals and nonpregnant female individuals who were 8 years and older.⁵

Multiple regression was used to predict BMI z score from weight z score and height z score. Percentiles and z scores are easily converted through the standard normal cumulative distribution function. A z score of 0 is the median, and the 97th percentile is 1.88 SDs above the median. Each SD increase in weight z score was associated with an increase in BMI z score by 1.15 SDs, whereas each SD increase in height z score was associated with a reduction in BMI z score of 0.54 SDs. Based on this model, which fit the data well (R2 = 0.94), a child at the CDC-defined 97th percentile for weight and height would have a predicted BMI z score of 0.01 + (1.15 × 1.88) − (0.54 × 1.88) = 1.16 SDs, the 88th percentile.

I also examined the amount of variability in body fat that BMI z score and WHD could explain among individuals aged 8 to 19 years. A combination of sex, age, and BMI z score yielded multiple R2 values of 0.88 (fat mass) and 0.76 (percent body fat), whereas the comparable R2 scores with WHD were 0.40 and 0.52, respectively.

A tall child whose weight is at the 97th percentile has a BMI above the 85th percentile because BMI is a height-adjusted weight index that is more strongly related to weight than height. In contrast, WHD assigns equal importance to weight and height. Despite its limitations, BMI z score predicts body fat more accurately than WHD.