Relative merits of the weight-corrected-for-height indices

Abstract

Based on weight and height data on 35,523 subjects representing both sexes, ages 18 yr and over, and five diverse ethnic populations, the obesity index W/H^p (p is derived from the observed weight-height data) was compared with the traditional obesity indices (W/H, W/H², W/H³, and H/W^1/3) in terms of maximum correlation with weight and minimum correlation with height (i.e., unbiased by height). Of the traditional indices, W/H and W/H² were highly correlated with weight, but W/H³ and H/W^1/3 were only moderately correlated with weight in each ethnic-sex-specific population. The indices W/H³ and H/W^1/3 were substantially more biased by height than were W/H and W/H² in each population. Of the two latter indices, W/H² was less height-biased in the male populations whereas W/H was less height-biased in the female population. However, all four indices, except W/H² in Chinese males, were significantly correlated with height (p < 0.05) in each population. The index W/H^p was highly correlated with weight and virtually independent of height in each ethnic-sex specific population. More importantly, W/H^p was also independent of height in an admixture of heterogeneous populations. This independence of height in pooled data makes the index invaluable for use in epidemiological and nutritional studies, where a single index must be used for comparing obesity among different populations. The relationship of each obesity index with weight and height did not materially change with age adjustment. The statistical results showed W/H³ and H/W^1/3 to be the least satisfactory indices of obesity, W/H and W/H² to be acceptable only in specific populations, and W/H^p to be decidedly the most suitable index in all populations. W/H^p should be routinely adopted as a weight-height-derived index of obesity.