OBJECTIVE: To determine the validity of self-reported body circumferen
ces and indices of body fatness in comparison with the same variables
measure by technicians. DESIGN: Cross-sectional survey. SUBJECTS: 66 w
omen aged 40-81 years. MEASUREMENTS: Self-reported weight and height a
nd waist, hip, chest and bust circumferences, by mail questionnaire, T
he same measurements taken by a trained technician during a clinic vis
it. Derived variables of body mass index, waist-to-hip ratio, and coni
city index, based on both self-reports and technician measurements. RE
SULTS: Mean differences between technician measurements and self measu
rements indicated that, on average, women tended to systematically und
erestimate their body circumferences. Age-adjusted Pearson correlation
s between technician measurements and self measurements ranged from 0.
93 (95% confidence interval: 0.89, 0.96) for hips to 0.99 (95% CI: 0.9
8, 0.99) for weight, For derived variables (quotients of measures), th
e highest agreement (r = 0.98, 95% CI: 0.97, 0.99) was observed for bo
dy mass index (weight in kg/height in m(2)), The ratio of circumferenc
es of the waist and hips was correlated less strongly (r = 0.76, 95% C
I: 0.63, 0.85), The correlation for conicity index, a measure of fat d
istribution that is independent of hip measurement variability, was 0.
82 (95% CI: 0.72, 0.88), Overall, accuracy of self measurements did no
t appear to Vary according to age. There was an indication that with i
ncreasing Values of weight and waist measurements, there was an increa
sing tendency for women to underestimate the measurement. This was als
o reflected in the accuracy of the derived variables body mass index a
nd waist-to-hip ratio, but not the conicity index. CONCLUSIONS:These r
esults indicate that self measurements of these anthropometric variabl
es are highly accurate even when used to formulate derived variables,
and are therefore appropriate for epidemiologic studies.