Classifying Weight Status Using Body Mass Index and Percent Body Fat

Abstract

Large scale studies use body mass index (BMI) to classify weight status (underweight, normal weight, overweight, obesity). BMI-defined weight status has demonstrated inconsistent relations with health outcomes. BMI may not differentiate between lean and fat tissue, which may confound relations with chronic disease morbidity and mortality. PURPOSES: 1) To determine relations between BMI weight status classifications and weight status classifications based on hydro static ally determined percent body fat (PFAT), which was considered the criterion. 2) To profile the inconsistencies between the two weight classification methods. METHODS: The study sampled 28,673 participants (Men=23,543; Women=5,130) who participated in a clinical evaluation of health and fitness at The Cooper Clinic (Dallas, TX) between 1970-2002. Clinical measures included height and weight from which BMI was computed as kg/m2 and PFAT from hydrostatic weighing. Participants were classified using BMI as not obese (< 30) or obese (> 30). PFAT was used to classify participants as not obese (Men: < 25%, Women < 33%) or obese (Men > 25%, Women > 33%). A sensitivity and specificity analysis was conducted to compare the two methods of health related weight classification. RESULTS: Based on PFAT, 7905 (28%) [Men = 6,878 (29%), Women = 1,027 (20%)] of the participants were classified as obese. A cross classification analysis (BMI × PFAT) indicated a sensitivity of 36% for BMI to correctly identify PFAT obese individuals with 64% false negatives. The specificity of BMI classifications was 96% with 4% false positives. CONCLUSIONS: There is a large degree of error in using BMI to identify obese individuals when PFAT is used as the criterion. BMI weight status does not effectively differentiate lean and fat tissue, which contributes to inconsistent relations with health outcomes. If obesity is an important risk factor, the large number of false negatives (obese who are classified as not obese) confounds any relative or attributable risk analysis of obesity with morbidity and mortality.