Figure 6

Association of body weight change (ΔBW) with omics modules. Shown are associations for the overall study population (column 1) and for subgroups (columns 2 to 11). Bubbles represent effect strengths and significance, as described in the legend. Models were adjusted for age, sex and baseline body weight. Significance threshold P <1.9 × 10⁻³ corresponds to Bonferroni correction for 27 modules. For subgroup analyses (columns 2 to 11), interaction models were fitted to obtain the main ΔBW effect in the respective subgroups, and the ΔBW:subgroup interaction effect indicating the difference in effect between the subgroups. Gene expression analysis was restricted to a subgroup of 689 subjects >55 years old at baseline, so that no effect estimates are available for the younger subgroup in this population (indicated as grey crosses). Note that all effects are shown per unit of ΔBW, which is a variable spanning the whole weight change range (with weight loss coded as negative ΔBW values and weight gain as positive ΔBW values). Thus, in the weight loss subgroup analysis, effects have to be inverted to obtain the effect per annual percentage body weight loss, and the same color of effects in the weight loss and weight gain subgroups denotes opposing effects of weight loss versus weight gain. GenM, gene expression module; MetM, metabolite module.