Fig. 3.

Estimates from multivariable observational analyses, one-sample MR analyses and two-sample MR IVW analyses for exposures BMI, WHR and WHRadjBMI per SD unit. Shown are point estimates, alongside 95% CIs, for the effect of each exposure on yearly hospital admission rate in UK Biobank participants of White British ancestry. The results are plotted on the log scale, to ensure symmetrical CIs and comparability of magnitude across estimates. Rate estimates and x-axis values are given on the rate scale. All Mendelian randomization analyses were adjusted for age, sex and the first 40 genetic PCAs. The multivariable observational analyses were adjusted for a range of baseline patient characteristics (Table 1). Notes: BMI = body mass index, CI = confidence interval, IVW = inverse variance weighted, MR = Mendelian randomization, SD = standard deviation, WHR = waist hip ratio, WHRadjBMI = waist hip ratio adjusted for BMI.