2018, Volume 10, Issue 1
Validation of accelerometer for measuring physical activity in free-living individuals
George O. Agogo1, Hilko van der Voet2, Paul J.M. Hulshof3, Pieter van ’t Veer3, Laura Trijsburg4, Fred A. van Eeuwijk2, Hendriek C. Boshuizen5
1Department of Biometrics, Wageningen University & Research, Dep. of Internal Medicine, Yale University, New Haven, USA
2Department of Biometrics, Wageningen University & Research
3Division of Human Nutrition, Wageningen University & Research
4Division of Human Nutrition, Wageningen University & Research, Nutritional Methodology and Biostatistics Group, International Agency for Research on Cancer, Lyon, France
5Dep. of Statistics, Mathematical Modelling and Data Logistics, National Institute for Public Health and the Environment (RIVM)
Author for correspondence: George O. Agogo; Department of Biometrics, Wageningen University & Research, Dep. of Internal Medicine, Yale University, New Haven, USA; email: email@example.com
Background: The aim of this research was to validate a triaxial GT3X accelerometer against doubly labelled water for measuring total energy expenditure (TEE) in a study of free-living Dutch adults and to compare the two prediction equations used to calculate accelerometerderived activity related energy expenditure.
Material/Methods: We used a measurement error model to estimate bias in the mean TEE, a correlation coefficient between measured and true TEE (a validity coefficient, which quantifies loss of statistical power to detect association) and the attenuation factor (which quantifies bias in the association), with and without conditioning on age, sex and BMI. We proposed
a calibration method for the accelerometer-based TEE.
Results: The accelerometer underestimated TEE by about 500kcal/day. The validity coefficient estimate conditional on age, sex and BMI was 0.8; the same value was observed for the attenuation factor estimate. With the devised calibration method, the bias in accelerometerderived mean TEE reduced to 6 kcal/day, validity coefficient estimate increased to 0.95 and attenuation factor to 0.94.
Conclusions: The GT3X accelerometer would underestimate mean TEE, lead to minimal loss in statistical power to detect significant association, and would result in biased estimate of the association between TEE and a health outcome.
Key words: attenuation, measurement error, physical activity, activity energy expenditure, validity coefficient