VALIDITY OF ESTIMATING MINUTE-BY-MINUTE ENERGY EXPENDITURE WITH ACCELEROMETRY

E.E. Kuffel 1, S.E. Crouter 2, J.D. Haas 2, E.A. Frongillo 3, D.R. Bassett 1, FACSM

1The University of Tennessee-Knoxville, Knoxville, TN 2Cornell University, Ithaca, NY 3University of South Carolina, Columbia, SC

PURPOSE : Researchers commonly use accelerometers to measure physical activity intensity and duration in free-living subjects, but these require regression equations to produce useful information. This study examined the accuracy of the Freedson, Hendelman, and Swartz regression equations and the Crouter 2-regression model for predicting metabolic equivalent values (METs) during transitions from rest to activity and back to rest.

METHODS: Thirty volunteers (age, 28±8 yrs) performed 15 minutes of supine rest, followed by 8 minutes each of: one-on-one basketball, seated rest, over-ground walking, and seated rest. An ActiGraph GT1M was positioned on the right hip and a portable metabolic unit was worn during all activities. Participants were randomly assigned to start the activity bouts at 0-s, 20-s, or 40-s into the minute (according to the ActiGraph clock).

RESULTS: The mean measured MET value for the basketball bout was 9.0 METs, compared to 7.7 METs for the Crouter 2-regression model, 5.4 METs for the Freedson equation, 5.0 METs for the Hendelman equation, and 6.0 METs for the Swartz equation (all, P<0.05). The mean measured MET value for the walking bout was 3.5 METs, compared to 3.9 METs for the Crouter 2-regression model, 3.7 METs for the Freedson equation, 4.1 METs for the Hendelman equation, and 4.6 METs for the Swartz equation (all, P < 0.05). The Crouter 2-regression model over-predicted measured METs during the first minute of all walking conditions and the last transitional minutes of the 20-s and 40-s walking conditions (P < 0.001).

CONCLUSION: During basketball the Crouter 2-regression model provided a more accurate estimate of METs than other commonly used single-regression ActiGraph equations. The tendency of the Crouter 2-regression model to over-predict the energy cost of walking results from the high variability in the accelerometer counts in the first and last transitional minutes.

This study was supported by NIH Grant 01R21 CA122430-01.