Evaluating Measures of Physical Activity and Sedentary Behavior Suitable for Large Epidemiologic Studies

Matthews CE, Keadle S, Moore SC, Troiano RP

ACSM 30 May – 3 June 2017, Denver; Med Sci Sport Exer 2017 May; 49 (5S) : 530

PURPOSE: Questionnaires typically used in epidemiologic studies to assess habitual physical activity (PA) and sedentary time (ST) are imprecise and do not assess the full spectrum of daily activities. This has limited our understanding of how these behaviors affect health. Thus, we tested two more precise and comprehensive measures, an accelerometer and an internet-based 24-hour recall (ACT24), as well as a questionnaire (Q). METHODS: Adults (50-70 y) enrolled in a 12-month study that included 3 criterion measures: PA energy expenditure (PAEE) measured by doubly labeled water (DLW), and two 7-d activPAL measures of active and sedentary time (A/ST). They also completed a 20-item past year Q about PA and ST (twice), an ACT24 recall of PA and ST every 2 months, and wore an ActiGraph (AG) on the waist twice for 7-d. AG equations (Freedson, Sasaki, Crouter 2-regression, Lyden machine learning [Soj3x]) were used to estimate PAEE. A/ST was quantified via Crouter, Soj3x and two cutpoints (100 vertical axis [100VA]; 200 vector magnitude [200VM]). Accuracy was evaluated as % difference from the mean of the criterion values, while correlations between the true criterion and each test instrument were also calculated. RESULTS: Among those with DLW or activPAL measures, mean age and BMI was 63 y and 23.8 kg/m2. Mean criterion values were: DLW PAEE (747 kcal/d, n=689) and activPAL A/ST (6.0/9.8 hrs/d, n=932). Accuracy in assessing PAEE was similar for the Q and ACT24 (-12% vs. 10%), but accuracy was lower for the Q than ACT24 for estimating A/ST (-48%/-18% vs. 3%/1%). Among AG measures, Crouter was most accurate for PAEE (-6%), while 100VA was most accurate for A/ST (1%/2%). Strength of the correlations for PAEE and A/ST were relatively consistent within each measure, but correlations were lower for Q (r=0.28-0.41), higher for ACT24 (r=0.63-0.73) and highest for AG Soj3x (r=0.72-0.77). Correlations for the other AG measures (r=0.61- 0.75) were similar to ACT24. Results did not vary greatly by sex, but all correlations for Q and ACT24 were lower for obese participants. CONCLUSION: ACT24 and AG measures of PA and ST were more accurate and had higher correlations than Q, suggesting that use of these more comprehensive measures in future epidemiologic studies could yield new etiologic discoveries, possibly leading to new intervention opportunities.

NA

Matthews CE, Keadle S, Moore SC, Troiano RP Evaluating Measures of Physical Activity and Sedentary Behavior Suitable for Large Epidemiologic Studies ACSM 30 May – 3 June 2017, Denver; Med Sci Sport Exer 2017 May; 49 (5S) : 530 2017. http://journals.lww.com/acsm-msse/Fulltext/2017/05001/Evaluating_Measures_of_Physical_Activity_and.1561.aspx.