LAUSR

www.thelancet.com Vol 391 May 5, 2018 1773 of BMI to define obesity (the degree of excess body fat) is highly specific, but has low to moderate sensitivity. As a result, BMI-based estimates of obesity prevalence are highly conservative for all ages and both sexes. Around half of all adults with excess body fat are defined as non-obese according to BMI, and at least 25–50% of children and adolescents defined as having a healthy BMI-for-age will also have excess body fat. These systematic reviews are based largely on evidence from European and western populations. In many other populations, the bias associated with the use of BMI is worse, leading to more pronounced underestimates of obesity prevalence, and complicating global comparisons. The inherent low sensitivity of BMI as a means of defining obesity should be acknowledged when prevalence estimates are provided. Informed comparisons between the prevalence of underweight and the prevalence of obesity can only be made if the large biases in obesity estimates are considered. Substantial underestimates of prevalence could also hinder future interventions aimed at obesity prevention and control. Obesity is far more common, and requires more urgent attention, than the NCD Risk Factor Collaboration study suggests.