Application and implications of new global definitions of obesity: a cross-sectional study of South African women

Background Recent guidance recommends body mass index (BMI) for obesity screening, with confirmation by direct body fat measurement or anthropometry, and introduces criteria for clinical obesity. The BMI threshold below 30 kg/m 2 at which confirmation should be applied is not specified. We aimed to operationalise the guidance in South African women by identifying the BMI threshold for confirmation and comparing obesity and clinical obesity prevalence and diagnostic performance across definitions. Methods In this cross-sectional study, anthropometry, body fat percentage (BF%) via air displacement plethysmography, and cardiometabolic markers were measured. Diagnostic performance of obesity definitions (BMI alone [≥30 kg/m 2 ], BMI with BF% confirmation and BMI with anthropometric confirmation) was evaluated using BF% as the reference standard. Obesity and clinical obesity prevalence were determined using each definition. Area under the curve (AUC) analysis informed selection of the BMI threshold below 30 kg/m 2 at which obesity confirmation should be applied. Findings Among 1379 women (median age 30 years), 63% had an elevated BF%. Overall, 52% (95% CI 50–55%) had a BMI ≥30 kg/m 2 , of whom 96% had elevated BF% and all had one or more elevated anthropometric marker. The optimal BMI threshold for predicting elevated BF% was 26.2 kg/m 2 , and we conducted obesity confirmation in women with BMI ≥25.0 kg/m 2 (n = 1208): 87% had elevated BF% (obesity prevalence in sample 60%, 95% CI 58–63%), and 93% met anthropometric criteria (obesity prevalence in sample 67%, 95% CI 64–69%). Sensitivity and specificity of BMI alone, BMI with BF% and BMI with anthropometry were 79% and 93%, 98% and 99%, and 96% and 75%, respectively. Among 1080 women with complete measures, 17% (95% CI 15–19%) and 18% (95% CI 15–20%) had clinical obesity using BMI with BF% and BMI with anthropometry, respectively. Interpretation Obesity confirmation increased prevalence estimates, primarily by extending assessment to below BMI 30 kg/m 2 . Anthropometric confirmation may overestimate obesity prevalence compared with BF% and improved measurement approaches are required in African settings.