Figure 2. pooling of longitudinal data: physical activity at baseline and risk of asthma incidence.

M-H; Mantel-Haenszel method, Random effects, CI; confidence interval. Not adjusted for potential confounders. Low physical activity used as reference category. Note that odds ratios are different of those in table 1 because reference categories were reversed and/or the number of categories was converted into two categories per study. For example Beckett et al. and Lucke et al. use high physical activity as reference category; in our meta-analysis we standardized low physical activity as reference category. In studies were more than two categories of physical activity were used (such as Beckett et al. who used 5 levels of physical activity), these were converted into two categories (in case of Becket et al. we converted the highest two levels into high physical activity, and the lowest three levels into low physical activity).