Agreement between BMI and body fat obesity definitions in a physically active population

• Main Author: Porto, Luiz Guilherme Grossi
• Publication Date: 2016
• Other Authors: Nogueira, Rosenkranz M., Nogueira, Eugênio C., Molina, Guilherme E., Farioli, Andrea, Junqueira Junior, Luiz Fernando, Kales, Stefanos N.
• Format: Article
• Language: eng
• Source: Repositório Institucional da UnB
• Download full: http://repositorio.unb.br/handle/10482/30498; http://dx.doi.org/10.1590/2359-3997000000220
• topic: Obesidade; Bombeiros; Índice de massa corporal (IMC); Gordura corporal

Body mass index (BMI) is a widely used proxy of body composition (BC). Concerns exist regarding possible BMI misclassification among active populations. We compared the prevalence of obesity as categorized by BMI or by skinfold estimates of body fat percentage (BF%) in a physically active population. Subjects and methods 3,822 military firefighters underwent a physical fitness evaluation including cardiorespiratory fitness (CRF) by the 12 min-Cooper test, abdominal strength by sit-up test (SUT) and body composition (BC) by BF% (as the reference), as well as BMI. Obesity was defined by BF% > 25% and BMI ≥ 30 kg/m2. Agreement was evaluated by sensitivity and specificity of BMI, positive and negative predictive values (PPV/NPV), positive and negative likelihood (LR+/LR-), receiver operating characteristic (ROC) curves and also across age, CRF and SUT subgroups. Results The prevalence of obesity estimated by BMI (13.3%) was similar to BF% (15.9%). Overall agreement was high (85.8%) and varied in different subgroups (75.3-94.5%). BMI underestimated the prevalence of obesity in all categories with high specificity (≥ 81.2%) and low sensitivity (≤ 67.0). All indices were affected by CRF, age and SUT, with better sensitivity, NPV and LR- in the less fit and older groups; and higher specificity, PPV and LR+ among the fittest and youngest groups. ROC curves showed high area under the curve (≥ 0.77) except for subjects with CRF ≥ 14 METs (= 0.46). Conclusion Both measures yielded similar obesity prevalences, with high agreement. BMI did not overestimate obesity prevalence. BMI ≥ 30 was highly specific to exclude obesity. Because of systematic under estimation, a lower BMI cut-off point might be considered in this population.