In Reply: In response to Dr Blumenthal and colleagues, self-selection is a generic problem of all population studies (and clinical trials) in which less healthy participants tend to decline participation. We tried to address the issue of residual confounding by adjusting for multiple factors, including ethnicity, education, smoking, depression, leisure activities, and comorbidities. Additionally, we conducted propensity modeling,1 which considered the probability of belonging to different physical activity groups conditional on individuals' background characteristics, thus creating participant characteristic balance between different physical activity groups (similar to a matching or a stratification process). With this modeling, the results were unchanged.
Scarmeas N, Luchsinger JA, Stern Y. Diet, Exercise Habits, and Risk of Alzheimer Disease—Reply. JAMA. 2009;302(22):2431-2432. doi:10.1001/jama.2009.1791