Validity of self-reported weight and height in the French GAZEL cohort

To examine the validity of self-reported weight and height and the resulting body mass index (BMI), and to explore the associations between demographic, socioeconomic, and health-related factors on the one hand and bias in self-reported weight and height on the other, in order to determine the groups most likely to exhibit bias.

Prospective cohort study.

7350 middle-aged subjects, 5445 men and 1905 women, from the GAZEL cohort, who have been followed up since 1989 and work at the French national company Electricité De France-Gaz De France (EDF-GDF) in various occupations.

Self-reported weight and height were based on information from yearly mailed questionnaires, and measured weight and height, used here as true values, were provided by occupational physicians from 1994 to 1997. Sex, age, marital status, education, occupation, history of ischemic heart disease, and treatment for cardiovascular risk factors were obtained from the mailed questionnaires or from data provided by the Company's personnel and medical departments.

Strong correlations were found between measured and self-reported values, but self-reported weight and height displayed significant systematic errors. Weight was significantly underestimated for men (0.54 kg) and for women (0.85 kg), and height overestimated for men (0.38 cm) and women (0.40 cm). These biases led to significant underestimations of BMI (0.29 and 0.44 kg/m2 for men and women respectively). Consequently, the prevalence of overweight, defined as BMI > 26.9 kg/m2 for women and BMI > 27.2 kg/m2 for men, was also underestimated, by 13% for men and 17% for women. The five factors associated with bias in self-reported weight and height were: overweight status, end-digit preference, age, educational level and occupation.

These findings suggest that self-reported weight and height should be treated with caution, because of biases leading to misclassification for overweight and obesity, especially in certain segments of the population.