Measuring Physical Activity in Free-Living Conditions—Comparison of Three Accelerometry-Based Methods

Abstract

We examined the agreement in time spent on different physical activity (PA) levels using (1) mean amplitude deviation (MAD) of raw acceleration from the hip, (2) wrist-worn Polar Active, and (3) hip-worn Actigraph counts using Freedson's cut-points among adults under free-living conditions. PA was measured in 41 volunteers (mean age 47.6 years) for 14 days. Two MET-based threshold sets were used for MAD and Polar Active for sedentary time (ST) and time spent in light (LPA), moderate (MPA), and vigorous (VPA) PA. Actigraph counts were divided into PA classes, ≤100 counts/min for ST and Freedson's cut-points for LPA, MPA, and VPA. Analysis criteria were simultaneous use of devices for at least 4 days of >500 min/d. The between-method differences were analyzed using a repeated measures analysis of variance test. Bland-Altman plots and ROC graphs were also employed. Valid data were available from 27 participants. Polar Active produced the highest amount of VPA with both thresholds (≥5 and ≥6 MET; mean difference 17.9–30.9 min/d, P < 0.001). With the threshold 3–6 MET for MPA, Polar Active indicated 19.2 min/d more than MAD (95% CI 5.8–32.6) and 51.0 min/d more than Actigraph (95% CI 36.7–65.2). The results did not differ with 3.5–5 MET for MPA [F(1.44, 37.43) = 1.92, P = 0.170]. MAD and Actigraph were closest to each other for ST with the threshold < 1.5 MET (mean difference 22.2 min/d, 95% CI 7.1–37.3). With the threshold <2 MET, Polar Active and Actigraph provided similar results (mean difference 7.0 min/d, 95% CI −17.8–31.7). Moderate to high agreement (area under the ROC curve 0.806–0.963) was found between the methods for the fulfillment of the recommendation for daily moderate-to-vigorous PA of 60 min. In free-living conditions the agreement between MAD, Polar Active, and Actigraph for measuring time spent on different activity levels in adults was dependent on the activity thresholds used and PA intensity. ROC analyses showed moderate to high agreement for the fulfillment of the recommendation for daily MVPA. Without additional statistical adjustment, these methods cannot be used interchangeably when measuring daily PA, but any of the methods can be used to identify persons with insufficient daily amount of MVPA.

Comparability of published cut‐points for the assessment of physical activity: Implications for data harmonization

“…Likewise, Figure shows large differences in the prevalence of our sample meeting the PA guidelines (ie, from 8% to 96%), so the fundamental query regarding the prevalence of the population achieving healthful levels of PA is still unresolved. In this regard, Leinonen et al found moderate‐to‐high agreement between different methods to classify adults meeting the PA guidelines. It is important to consider that PA guidelines have been developed predominantly using self‐reported data, thus, these estimations should be considered with caution.…”

Associations of physical activity in detailed intensity ranges with body composition and physical function. a cross-sectional study among sedentary older adults

“…Due to the very limited amount of vigorous-intensity activity, moderate and vigorous intensity activities were combined. The cut-points have been defined and validated against VO 2 [25,26], and compared with widely used Freedson's cut-points for activity counts from uniaxial ActiGraph GT3X [27] in healthy younger adults, but not in older adults. In the last-mentioned study, MAD values showed slightly more sedentary activity, but notably less light activity and more moderate activity than activity counts.…”

“…In the last-mentioned study, MAD values showed slightly more sedentary activity, but notably less light activity and more moderate activity than activity counts. The amount of vigorous activity was similar [27]. To investigate the physical activity intensity range in detail, a second histogram with histogram bins from zero to 1.2 g in base 10 logarithmically equidistant bins was calculated [28], which resulted in 93 bins with at least some activity recorded.…”

Prolonged bouts of sedentary time and cardiac autonomic function in midlife

“…In our earlier comparison of different accelerometry-based methods, the threshold <2 MET for Polar Active provided similar results as threshold <100 cpm for ActiGraph, and the mean difference between methods was 7.0 minutes/d (95% confidence interval from −17.8 to 31.7 minutes/d). 34…”

“…The causality of the findings cannot be concluded because of the cross-sectional study design. The PA monitor might overestimate the amount of vigorous PA, which was in this study combined with moderate PA. 34 The lack of posture recognition in measuring SED with wrist-worn monitors is also a limitation as the physiological consequences of sitting and standing can differ, a recent study reporting higher levels of occupational standing to be associated with higher rMSSD. 25,40 The reliability of short-term HRV recording has been argued, although its advantage in terms of fast measurement, data analysis, and low participant burden has been acknowledged.…”