Estimates of the sources of variation (variance components) of bioelectric impedance and anthropometric measurements in an epidemiological case-control study of breast cancer

Objective: This study investigated the variability of anthropometric measurements and body fat estimated by bioelectric impedance analysis. Subsequently the methods were applied in a case-control study to investigate the association with breast cancer. Design, Subjects: The study group included 50 consecutive cases and 75 age-matched controls from the same area The variation was investigated in 50 healthy women from the control group, who were repeatably measured using standardised measurement procedures, and the variation between-subjects, within-subjects, between-observers, and within-observers were estimated. Results, Conclusions: The study showed that the variance components between-subjects were 64-99% of the total variance. The variables of skinfold thicknesses were characterised by having the highest relative observer variation and having many unavailable values that were out of the range of the Harpenden callipers. The mean body fat by bioelectric impedance analysis was 31.2%, and the total coefficient of variation 23%, while the variance components related to subject time, observer and measurement were 98.4%, 1.1%, 0.2%, and 0.4%, respectively. The body fat was significantly correlated with the variables of skinfold thicknesses. We decided to exclude the variables of skinfold thicknesses from the case-control study, and for the other variables to measure each subject only at one time by one observer. The case-control part of the study indicated a non-significant increase in body weight in the postmenopausal breast cancer patients (mean difference 3.6 kg; confidence interval from -0.9 kg to 8.0 kg). Similarly the body fat tended to be higher in the breast cancer patients (mean difference 1.2%; confidence interval from -1.6% to 4.0%).