Alternative measurements of obesity: accuracy of body silhouettes and reported weights and heights in a Mexican American sample.

Abstract

The assessment of obesity by anthropometry or body composition in studies of chronic disease epidemiology is not always feasible. In this paper we test the accuracy and validity of two alternative methods: body silhouettes and reported weights and heights in a sample of Mexican American adults (n = 166) participating in the Diabetes Alert Study. The body silhouettes were those developed by A. Stunkard and colleagues. We compared the silhouettes obtained independently by three different observers, one with minimal experience in assessing obesity, the other two skilled in anthropometry. Correlations between the expert observers were 0.89-0.90. Those between the expert and less skilled observers were lower (0.64 to 0.85) and were better for female than male subjects. Results suggest an acceptable level of precision for persons with some training in obesity assessment. Body silhouettes compared well with the body mass index (r = 0.85 to 0.92 for the expert vs r = 0.65 to 0.84 for the less skilled observer), being higher for female subjects. The body silhouettes are thus useful for categorizing normal, overweight and obese individuals. The poorer performance on male subjects may relate to the female-like obesity depicted in both sexes in this particular set of silhouettes. A substantial proportion of subjects could not recall their weight or height. For those who could recall their measurements, accuracy was good but measurement error higher than that for measured variables. These problems of recall in this sample from rural Texas point to the potential usefulness of the body silhouette method when actual height and weight measurements cannot be made.