Contents - Previous - Next

An assessment of nutritional factors affecting the BMI of a population

Introduction
Methods
Results
Discussion
References

P. J. François¹ and W. P. T. James²

¹Formerly of FAO Nutrition Division (current address: 12 Square pont de Sevres, Paris 92100, France); and ²Rowett Research Institute, Greenburn Road, Bucksburn, Aberdeen AB2 9SB, UK

This paper assesses some of the environmental factors relating to BMI in a very large and comprehensive household survey conducted in 1974 in Brazil and involving about 51000 households in all parts of the country. In this study not only were the children and adults in all households weighed and measured but the household diet, socio-economic status and a variety of issues relating to household practices were assessed as part of a major national economic survey conducted by the Brazilian Institute of Geography and Statistics with technical support from FAO. One of the authors (P.F.) was the principal collaborator from FAO.

Introduction

It is well recognized that within any society the size of a child or adult reflects the outcome of interactions between environmental factors and the genetic potential for growth. The assumption that national differences in the growth of children was dependent on genetic differences has given way to a recognition that in most Third World populations infection and diet have the dominant roles in explaining national differences in children's growth rates. Within any group of children, however; the final attained height does reflect the contribution of individual differences in growth potential.

In adults, the determinants of changes in body size have been equally debated. Thus, studies on obesity have either emphasized the impact of genetic susceptibility to obesity or the psychological disturbance leading to or developing in patients with severe forms of obesity. The genetic contribution to body mass indexes (BMIs: kg/m²) within a population has been considered to range between 25% and 75% of the variance in BMI (Bouchard & Tremblay, 1990; Sorensen, Stunkard & Holst, 1991). Much less attention has been paid, however, to the dietary and other factors, such as physical

Correspondence to: Professor W. P. T. James. activity, which determine either the average BMI of a population or the distribution of BMIs within a society.

Methods

Each householder was visited during the preparation of each meal on each of 7 days in a week's monitoring of the household. Daily food use, edible food production and discarded food was weighed for the household as a whole with 1500 different food items being recorded in the survey. Attendance at meals was also recorded. A total of about 300000 individuals in 51360 households were involved.

The large data sets were first assessed for the dominance of different food patterns by multivariate factor analysis. The approach to this type of analysis to establish the major common features in a huge variety of different patterns of response has been described by Hill (1974) and Lebart, Morineau & Warwick (1984). The dietary patterns were also assessed not only as foods but also in nutritional teens by relating the estimated protein, fat, carbohydrate and alcohol content to the total energy of the household diet. In the discrimination of different diets, 13 classes were chosen for the range of protein/energy ratios within the diets recorded with 20 further classes each for the fat and carbohydrate energy ratios (10 each) and five classes for alcohol consumption. Ten categories of income and 10 of BMI were also chosen making a total of 68 classes of six variables. These classes were then used to characterize the dietary pattern of all 52 340 adult males surveyed. Each individual could obviously occupy only one class for each variable. By accumulating the total scores of individuals for each class in each variable, an individual profile was obtained for all 52340 men. The profiles of these men could then be distributed in a scatter plot which took account of the six variables and where individuals with the same profile end up with the same coordinates. The distance between the points on a scatter diagram then signified the x2 difference between the variables, the dominant effects producing a concentration of points appropriate to the particular variables. By extracting the dominant variables in the spread of data one can then set out a two-dimensional diagram based on the two principal but independent functions.

Table 1. The nutrient intake and body mass index (BMI) of adult males in different incrome groups in Brazil in 1974 classified by three dietary patterns classified by staple consumption

This very crude portrayal of the process allows the non-statistician to obtain some understanding of the process of analysis but any mathematically-orientated reader is advised to consult Hill (1974) and Lebart et al. (1984).

Contents - Previous - Next