BACKGROUND: Since subjects included in population studies tend to underreport their weight and overestimate their height, obesity prevalence based on these data is often inaccurate. A reduced obesity threshold for self-reported height and weight was proposed and evaluated for its accuracy. METHODS: Self-reported heights and weights were compared with measured heights and weights in a Swiss city adult population representative sample. Participants were asked their height and weight and were invited to undergo a health examination, during which these data were measured. An optimal body mass index (BMI) value was assessed using receiver operating characteristic (ROC) curve analysis and its ability to correctly estimate obesity prevalence was tested on an external French population sample. RESULTS: The Swiss population sample consisted of 13 162 subjects (mean age 51.4). The comparison between self-reported and measured data showed that obesity prevalence calculated from declarations was underestimated: among obese subjects (according to measured BMI), 33.6% of men and 27.5% of women were considered to be non-obese according to their self-report. Considering measures as a reference, a lower BMI cut-off of 29.2 kg/m(2) was identified for both genders for the definition of obesity based on self-report. Respective misclassification was reduced to 17.9% in men and 16.9% in women. The validation procedure on a French population sample (n = 1858) yielded similar results. CONCLUSIONS: The reduced threshold based on self-report allowed a better estimation of obesity prevalence. Its use should be limited to population studies only.
BACKGROUND: Since subjects included in population studies tend to underreport their weight and overestimate their height, obesity prevalence based on these data is often inaccurate. A reduced obesity threshold for self-reported height and weight was proposed and evaluated for its accuracy. METHODS: Self-reported heights and weights were compared with measured heights and weights in a Swiss city adult population representative sample. Participants were asked their height and weight and were invited to undergo a health examination, during which these data were measured. An optimal body mass index (BMI) value was assessed using receiver operating characteristic (ROC) curve analysis and its ability to correctly estimate obesity prevalence was tested on an external French population sample. RESULTS: The Swiss population sample consisted of 13 162 subjects (mean age 51.4). The comparison between self-reported and measured data showed that obesity prevalence calculated from declarations was underestimated: among obese subjects (according to measured BMI), 33.6% of men and 27.5% of women were considered to be non-obese according to their self-report. Considering measures as a reference, a lower BMI cut-off of 29.2 kg/m(2) was identified for both genders for the definition of obesity based on self-report. Respective misclassification was reduced to 17.9% in men and 16.9% in women. The validation procedure on a French population sample (n = 1858) yielded similar results. CONCLUSIONS: The reduced threshold based on self-report allowed a better estimation of obesity prevalence. Its use should be limited to population studies only.