Three limitations of the body mass index

Abstract

Listen

Translate article

Although the wt/ht2 or Body Mass Index (BMI) has attained growing popularity in the past few years (1), the two components of this index and their relative weighting suggest possible limitations on its use. Since stature (ie, standing height) is one component, the BMI may be stature dependent over at least part of the age range. Again, the use of stature as a divisor suggests that the BMI may also be a!fected by relative leg length or relative sitting height. Finally, the use of weight as the numerator suggests that the BMI-like weight itself-may reflect both lean tissue and fat tissue to a comparable degree. If we turn first to the extensive data of the first National Health and Nutrition Examination Survey (NHANES I), readily available on tape and familiar to most readers of this journal (2), the BMI is indeed stature dependent but in different directions at different points in the life cycle. For NHANES I, correlations between stature and the BMI approximate 0.30 for children, shift during adolescence, and become negative in adult years (Table 1). The massive data of the National Collaborative Perinatal Project (NCPP) confirm these trends; correlations between stature and the BMI again approximate 0.30 for children but average -0. 12 for 40 000 3rd and 4th decade women. Thus the assumption that the BMI is independent ofstature is not quite true for adults and especially not true for children. Anthropometric data included in NHANES I also reveal the extent to which BMI reflects body proportions or more specifically sitting height relative to standing height (Table 2). At all ages considered (ages 5-50) the correlation between BMI and relative sitting height (sitting height/stature) is at least 0.15 and generally exceeds 0.20. With n = 596 for adolescent males to n = 3858 for young adult women, all ofthese values of r are statistically significant. However, the point in question is the magnitudes of the correlations and their implications. Children, adolescents, or adults with short legs for their height have higher BMI values, which indicates the extent to which BMI is also a measure of body build or body proportions. When the BMI is further explored in relation to frame size, Lean Body Mass, and fatness, again on an age-specific basis, it is seen to relate to all three. With use of data from the Tecumseh (Michigan) Community Health Survey to include the radiogrammetricallydetermined bony chest breadth (3) as in Table 3, the correlations approximate 0.5 overall. For the same subjects arranged by age intervals, correlations with the triceps skinfold approximate 0.7, as shown in the table. If we further estimate Lean Body Mass by subtracting estimated fat weight from total body weight (4), correlations with BMI then approximate 0.65 for males aged 30-50. Accordingly, it is clear that BMI reflects both the weight of lean tissue and the weight offat tissue and for some age groups, at least, it may be a better measure of amount oflean than of relative fatness. Under these circumstances and with sampies of considerable size (up to and exceeding 40 000), it is clear that the BMI is not quite independent ofstature, especially at the younger ages. Moreover, the BMI is influenced by