Adolphe quetelet and the BMI: Fact, fiction, and childhood growthписьмо
Аннотация: It is often stated, with or without supporting references, that the BMI was proposed by Adolphe Quetelet (e.g., see Refs. 1,2). However, I find no evidence for this in his own writings, of which I cite only three here ((3-5)). What he did assert was that body mass in adults tends to vary with the square of height ((3-5)). Less often noted is that he also wrote that mass2 increases with height5 in adolescents ((3-5)), implying that a more appropriate version of the BMI for them is not mass/height2, but mass/height2.5. This accords approximately with values of p in the Benn Index (i.e., mass/heightp) tabulated in a notable paper by Cole for successive one-year age groupings ((6)). (That height and mass were standardized for age is unimportant here.) Estimates of p were generally 2.2-3.3 in children between four and about 15 years of age. The highest values tended to occur around the time of puberty. None significantly exceeded 3.0, which is the value Quetelet recognized as appropriate to isometric growth ((3-5)). The variation in p largely relates to variation in the correlation coefficient, rMH, for the logarithms of mass-for-age and of height-for-age. Cole's Table 2 reveals a marked correlation between the values of p and rMH in both boys and girls from the age of six (n = 40; r = 0.89; P < 10−10). This recalls a similar correlation between estimates of p and the respective correlation coefficients for mass and height in adults ((7)). In the children, again from the age of six, the 40 values of rMH correlated with the standard deviations of log(height-for-age) (r = 0.55; P < 0.0002). Thus, in those age groups in which the adolescent growth spurt makes height most variable, rMH tends to be highest and p tends to be closest to 3. The correlations between p and rMH in adults and children are inherent in the statistical techniques used in estimating p ((7)). With adequate samples, the estimates that they produce will always be less than the true (functional) values. It means that Quetelet's assertion and continuing common belief, that adult body mass tends to vary with height2, is actually wrong (though this does not invalidate the BMI as a predictor of %adiposity in adults). The apparent allometry in adults is largely or entirely a statistical artifact. The childhood peak in p is also largely or entirely a statistical phenomenon, explainable in terms of growth, but not of changing bodily proportions such as might relate to adiposity. The statistical reasoning is uncommon in anthropometry and can seem counterintuitive, but it is readily tested by Monte Carlo modeling ((7)). Richard F. Burton*, * School of Life Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow UK
Год издания: 2012
Авторы: Richard Burton
Издательство: Wiley
Источник: Obesity
Ключевые слова: Medicine and Dermatology Studies History, Neurology and Historical Studies, Medical History and Innovations
Открытый доступ: bronze
Том: 21
Выпуск: 1
Страницы: 6–6