To the editor The recent review of Hwaung et al provides a detailed and valuable account of the history and rationale of the body mass index (BMI) and of indices relating waist circumference (WC) to body height that take the form WC/height α . 1 I show here how their evidence and analysis are usefully supplemented with those of earlier studies.The ratio WC/height, with α = 1, is widely used, but the authors have concluded that the optimum value of α is approximately 0.5, with this giving the strongest association with adiposity and the weakest correlation with height for men and women of four race/ethnic groups.Burton 2 also considered 0.5 to be an appropriate round-number value.On dimensional grounds, one might choose a value of 1 for α, making the index dimensionless, but data scatter associated with variable body shape must lower values of α as estimated by regression analysis-just as it lowers the height exponent, p, of the Benn index, (body mass)/height p . 2,3 (Here, I am applying the symbol "α" just to WC and not also to body mass as in the review.) Both α and p necessarily correlate strongly with the respective correlation coefficients for WC and height and for body mass and height. [2][3][4] The negative value of α for Korean women (−0.43)-unadjusted for age-must be associated with a negative correlation between WC and height (though R 2 is positive).These negative values were also found by Han et al for Europeans. 4 This, as well as Table 1, 1 illustrates the importance of age in considering the height dependences of WC and body mass and therefore both α and p. In contrast, Table S9 indicates that there is conveniently little influence of age on correlations between WC and % fat and between body mass and % fat. With increasing age, body mass tends generally to increase and height tends to decrease. 4 Figure 2 of Burton 2 shows a strong correlation between p and α (with α there denoted q) with most of the points being for data that were age-adjusted or grouped for age. 4 The tabulated results of Hwaung et al show similar relationships, for values both unadjusted and adjusted for age. 1 Moreover, their eight age-adjusted values, together with eight age-adjusted values of Han et al., 4 show a single clear straight-line relationship, with a correlation coefficient of 0.91. This suggests a link beyond the obvious between WC/height α and (body mass)/height p that has yet to be elucidated. The reduced major axis regression equation is: α ¼ 0:69p − 0:78: Hwaung et al discussed the question of whether 1/BMI and 1/ (WC/height α ) are additive in multiple regression models for the prediction of % fat. Their table S7 shows, for age-corrected values, that, for men, R 2 for % fat and 1/(WC/height 0.5 ) is higher than for % fat and 1/ BMI, while the opposite is true for women. Multiple regression of % fat on both indices together did not increase R 2 above the highest of the two values by more than 0.01. So the two indices are not usefully additive. As a correlate or predictor of % fat, the better index evidently d...