Being overweight is a risk factor for general health outcomes as indicated by its risk effect on disease count (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat b_{xy} = 0.41$$\end{document}b^xy=0.41) in the community data. The question is then how bxy for disease count should be interpreted. We have shown in Supplementary Fig. 9 that the estimate of bxy for disease status (a dichotomous phenotype to indicate whether an individual is affected by any of the 22 diseases) was very similar to that for disease count. Although disease status and disease count are two distinct phenotypes and the analysis of disease count is more powerful, for the ease of interpretation, bxy for disease count can be approximately interpreted as logOR for disease status. Hence, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat b_{xy} = 0.41$$\end{document}b^xy=0.41 for disease count is approximately equivalent to OR = 1.51 for disease status, meaning an increase of BMI by 1 SD will increase the probability of being affected by any of the 22 diseases by a factor of ~1.5. In addition, we found that the effects of WHRadjBMI and BMI on disease were largely concordant (Supplementary Fig. 10a; Supplementary Note 4).
