Classification of the degree of obesity by body mass index or by deviation from ideal weight

Abstract

BACKGROUND: The purpose of this study was to compare classifications of subjects as underweight, normal weight, or obese by body mass index (BMI) and the ratio of body weight to ideal weight (W/IW). METHODS: We performed a theoretical comparison of the 2 indices. We compared classifications of the degree of obesity in 1839 women and 5914 men who were followed up in the primary care clinics of a United States federal hospital. Information was extracted from computerized records. Subjects were classified as underweight (BMI < 18.5 kg/m2, W/IW < 0.9), obese (BMI > or = 30.0 kg/m2, W/IW > or = 1.2), or normal weight (BMI, W/IW values between the cutoff values for underweight and obesity). W/IW values were computed assuming small, medium, and large skeletal frame for all. We compared the classifications of subjects as underweight, normal weight, or obese by BMI and W/IW. We used Cohen's kappa ratio to evaluate the agreement between these classifications. RESULTS: Theoretically, the cutoff values of BMI and W/IW for underweight and obesity are not in agreement. Patient data revealed substantial differences in the classifications of subjects as underweight, normal weight, or obese. Kappa ratios ranged between 0.18 (poor agreement) and 0.71 (reasonable, but not high degree of agreement). In general, kappa ratios were higher when assuming large or medium skeletal frame versus small frame. CONCLUSIONS: There are substantial discrepancies in classifying the subjects of a population as underweight, normal weight, or obese by BMI or W/IW. These discrepancies may cause confusion when 2 or more indices are used simultaneously to classify the degree of obesity.

To identify potential discrepancies of classification of the degree of obesity by different height-weight indices, we compared the formulas defining body mass index and ideal weight and the classification of 1839 women and 5915 men as underweight, normal weight, or obese by the 2 height-weight indices.