Charles Matthews

NIH

Senior Investigator

IDATA (Learn more about this study)

IDATA-9

Feb 13, 2017

Evaluation of Physical Activity and Sedentary Behavior Measures Suitable for Large-scale Epidemiologic Studies

Physical activity questionnaires designed to examine the relation between behavior-disease associations in large-scale epidemiologic studies are highly demanding from a cognitive perspective—relying on long-term recall and estimation of uncertain autobiographical information. Furthermore, most questionnaires have not been designed to capture the full range of sedentary and physical activity behaviors, and thus provide an incomplete picture of certain aspects of human behavior. The cognitive demands inherent in our questionnaire-based assessments result in significant exposure measurement error, and in prospective studies such errors typically result in an attenuation of the magnitude of observed associations, a loss of statistical power, and can increase the rate of type II errors (false negatives) in our studies. Furthermore, the omission of many potentially important physically active and sedentary behavior behaviors from most questionnaires is a major impediment to investigation of the health effects of these exposures and/or patterns of behavior. Limitations in our questionnaire-based measures are a fundamental barrier to our understanding of whether physical activity and sedentary behavior may prevent certain cancers or improve prognosis following a cancer diagnosis, and for translating our etiologic findings into clinical and community-based prevention efforts.

In recent years, previous-day recalls have emerged as a less error prone approach for measuring physical activity and sedentary behaviors by self-report, and accelerometers facilitate the accurate and precise measurement of these behaviors in a manner than is completely free of reporting error. In addition, these new “better” measures also assess a more complete range of daily activities and provide insight into patterns of behavior that may influence health, information typically unavailable from traditional questionnaires. Thus, previous-day recalls and accelerometers provide an opportunity to improve on traditional questionnaire-based measures in large-scale epidemiologic studies and offer the opportunity to advance our understanding of the role that physical activity can play in cancer prevention and control.

The iDATA Study was designed to evaluate a wide range of measurements of physical activity and sedentary behavior suitable for large-scale prospective studies of cancer risk and overall health. To advance the science in this area of inquiry, we propose the following research objectives.

Objectives
1. Describe the amount, type, and patterns of physical activity and sedentary behaviors in adult AARP members, as well as the behavioral variability in key summary metrics in the Test Measures and Reference Measures obtained in the study (described below).

2. Describe the measurement properties of the Test Measures in comparison to the Reference Measures. A variety of measurement properties of the Test Measures will be evaluated (e.g., R2, root mean square error (RMSE), bias, the slope of the relation between measures, and systematic and random errors) using simple statistical tests, linear regression, and measurement error models. We will also evaluate differences in measurement properties of the Test Measures by participant characteristics and a variety of methodologic decisions/ assumptions.

3. Using results from Objective 2, apply measurement error correction methods for physical activity and sedentary behavior to behavior-disease associations in the NIH-AARP Diet and Health Study.

4. Describe the impact of patterns of behavior in available measures on relevant public health and etiologic metrics (e.g., duration of moderate-vigorous physical activity, sedentary time, physical activity energy expenditure).

Charles E. Matthews, PhD (DCEG, Metabolic Epidemiology Branch)
Stephen C. Moore, PhD (DCEG, Metabolic Epidemiology Branch)
Pedro Staint-Maurice, PhD (DCEG, Metabolic Epidemiology Branch)
Joshua N. Sampson, PhD (DCEG, Biostatistics Branch)
Sarah Kozey Keadle, PhD (CalPoly)
Raymond Carroll, PhD (Texas A&M)
Dale Schoeller, PhD (Univ Wisconsin)
Richard P. Troiano (DCCPS, Risk Factor Assessment Branch)