Pedro Saint-Maurice

Ph.D.

NCI

Postdoctoral fellow

IDATA (Learn more about this study)

IDATA-12

Jun 7, 2017

REPRODUCIBILITY OF PHYSICAL ACTIVITY MEASURES IN ADULTS AND IMPLICATIONS FOR EPIDEMIOLOGICAL STUDIES

Physical activity (PA) is a multidimensional construct that requires precise measurements to be appropriately quantified. Physical activity questionnaires are the most popular measure of PA however objective measures such as accelerometers are known to produce more precise estimates of PA. Despite the potential of accelerometers for measuring PA these tools have evolved over time to fill the increased demand for precision in measuring PA and in examining the relations between PA and disease outcomes (e.g., cancer, mortality). Recently established epidemiological cohorts (e.g., UK BioBank) have now incorporated accelerometer measures however, this effort is still compounded by both logistic, and economic challenges. The incorporation of accelerometer devices in PA measurement protocols for large-scale epidemiological studies requires a considerable economic investment and a comprehensive staff team in order to collect and process this type of data. For these reasons, PA measurement protocols have historically relied on a single-7 day (i.e., 1 week) assessments to characterize activity.

Much research has examined on “how many days” of assessment are needed but only for the period measured. Instead, very few studies have examined how well one 7-day assessment captures individual PA variability in adult populations. Studies have shown that there is considerably variability in day-to-day PA and that activity patterns are also likely to vary across seasons. This variability is known to increase measurement error associated with measures obtained from accelerometers which is expected to attenuate estimates of risk in etiological studies interested in PA as the main exposure. The increased susceptibility of these measures to error has typically been addressed at the study design stage by increasing the sample size of participants being studied. A less frequent strategy to handle measurement error has also been to increase the number of PA assessments by using replicate measurements throughout a longer period of time (e.g., 2 assessments over the course of 1-year). Each of these strategies have important implications for study design and both require a good understanding of variability associated with PA behavior over time. The current proposal addresses this gap by examining the stability of PA measures over a 6-month period in adults and its implications for designing epidemiological studies.

1. Examine variability (i.e., measurement error) in estimates of PA and sedentary behavior resultant from both criterion and surrogate measures of PA. We are particularly interested in determining the reliability of PA energy expenditure, minutes spent in intensity-specific PA (i.e., sedentary, light, moderate, and vigorous PA), and postural-related metrics obtained from ActivPAL (i.e., sitting, stepping, walking). Reliability will be determined by computing Intraclass Correlation Coefficients (ICCs) and respective 95% confidence intervals (95% CI).

2. Determine the implications of measurement error associated with each metric listed above for determining the number of replicate measurements in large-scale epidemiological studies. The resultant ICCs from Aim 1 will be used to quantify the number of replicate assessments and understand viable options to minimize this type of error.