Evaluating the direct effects of childhood adiposity on adult systemic metabolism: a multivariable Mendelian randomization analysis

Figures

• 

  Figure 1.. Schematic causal graphs used to illustrate (A) univariable Mendelian randomization (MR) analyses used in this study to estimate total effects between childhood body size and circulating metabolites, (B) multivariable MR analyses to estimate direct effects of childhood body size of circulating metabolites and (C) applying the same multivariable framework to estimate the indirect effects on circulating metabolites mediated along the causal pathway via adult body size. The highlighted arrows on these graphs illustrate the causal effect of childhood body size on the outcome being estimated in MR analyses. The textured arrows and grey shading indicate the effects that MR is typically robust to in comparison to observational analyses.

• 

  Figure 2.. Receiver operator characteristic (ROC) curves to compare the predictive ability of genetic scores for childhood body size (blue) and adult body size (red). (A) ROC curve to investigate the prediction of adiposity during childhood (N = 2427, age = 3–18 years) using cut-offs defined in Supplementary Table 4 (available as Supplementary data at IJE online) and (B) ROC curve to investigation prediction of adiposity during adulthood based on BMI ≥ 30 kg/m2 (N = 1762, age = 34–49 years).

• 

  Figure 3.. A forest plot depicting the observational (orange) and genetic (green) effect estimates between childhood body size (per change in body size category) and circulating metabolites (per standard deviation unit change). Observational estimates were derived using data from the childhood time point from the Young Finns Study, whereas genetic estimates are based on two-sample Mendelian randomization (MR) analyses using summary data. The observational estimates in this figure have been scaled using a scale factor to have the same magnitude of dispersion around the central estimates as the MR results for comparative purposes.

• 

  Figure 4.. A forest plot illustrating the effect estimates of genetically predicted childhood (yellow) and adult (blue) body size (per change in body size category) on circulating metabolites (per standard deviation change) based on multivariable Mendelian randomization analyses. Points corresponding to estimates whose confidence intervals overlapped with the null were not filled in.

• 

  Figure 5.. A forest plot illustrating the effect estimates of genetically predicted metabolites (per change in standard deviation increase) on coronary artery disease risk based on univariable Mendelian randomization analyses. Points corresponding to estimates whose confidence intervals overlapped with the null were not filled in.