Isotemporal substitution model: a novel statistical approach for physical activity epidemiology

Abstract

An individual’s discretionary time in a given day is finite, and the components (i.e., behaviors) are allocated in an interdependent manner. A novel statistical approach named the “isotemporal substitution (IS) model” that was developed to consider this interdependency has recently gained popularity, and shows various advantages over typical analytic models used in the field of physical activity epidemiology. Despite these advantages, there have been no original or introductory articles on the IS model emerging from Japan. The purposes of this review were to provide an overview of the definition, assumptions, and advantages of the IS model; to summarize the existing literature; and to explore research perspectives in this field. The IS model was designed to estimate the effects of replacing one behavior with another for the same amount of time on outcome measures. This is accomplished by entering a total combined behavior variable (i.e., time spent in sedentary, light, and moderate-vigorous activities, corresponding to the wear time of an accelerometer) along with specific behavioral variables (except for the behavior variable of interest) into the model at the same time. The model then statistically holds the wear time constant, drops the target behavioral variable being reduced, and then enables estimation of the “substitution” effects of reallocating one behavior to another on the outcome measures (e.g., the effects of substituting 60 min/day of sedentary time with moderate-vigorous physical activity). The greatest advantage of the IS model is its better interpretability compared to typical models. This has a potential benefit for establishing public health recommendations and health promotion practices. As of July 29, 2015, a systematic literature search identified 12 relevant articles utilizing the IS model. The content review revealed research needs for studying activity type and posture allocation as exposure variables, clinical populations as subjects, and a longitudinal design. These themes should be future research priorities. We anticipate active use of the IS model and its widespread applications in physical activity epidemiology in Japan.