Within- and between-laboratory precision in the measurement of body volume using air displacement plethysmography and its effect on body composition assessment

OBJECTIVE: To determine and compare the extent of within- and between-laboratory precision in body volume (BV) measurements using air displacement plethysmography (ADP), the BOD POD body composition system, and to interpret any such variability in terms of body composition estimates. DESIGN: Repeated test procedures of BV assessment using the BOD POD ADP were reproduced at two laboratories for the estimation of precision, both within and between laboratories. SUBJECTS: In total, 30 healthy adult volunteers, 14 men (age, 19–48 y; body mass index (BMI), 19.7–30.3 kg/m 2 ) and 16 women (age, 19–40 y; BMI, 16.3–35.7 kg/m 2 ), were each subjected to two test procedures at both laboratories. Two additional volunteers were independently subjected to 10 repeated test procedures at both laboratories. MEASUREMENTS: Repeated measurements of BV, uncorrected for the effects of isothermal air in the lungs and the surface area artifact, were obtained using the BOD POD ADP, with the identical protocol being faithfully applied at both laboratories. Uncorrected BV measurements were adjusted to give estimates of actual BV that were used to calculate body density (body weight (BWt)/actual BV) from which estimates of body composition were derived. The differences between repeated BV measurements or body composition estimates were used to assess within-laboratory precision (repeatability), as standard deviation (SD) and coefficient of variation; the differences between measurements reproduced at each laboratory were used to determine between-laboratory precision (reproducibility), as bias and 95% limits of agreement (from SD of the differences between laboratories). RESULTS: The extent of within-laboratory methodological precision for BV (uncorrected and actual) was variable according to subject, sample group and laboratory conditions (range of SD, 0.04–0.13 l), and was mostly due to within-individual biological variability (typically 78–99%) rather than to technical imprecision. There was a significant ( P