Fig. 2.

A Orchard plots showing mean effects of increased dietary BCAAs on plasma levels of BCAAs (total and individual). Thick error bars are 95% confidence intervals (CI; i.e. statistical significance) and fine error bars are 95% prediction intervals (i.e. heterogeneity in reported effects; the range within which 95% of effect sizes are expected to fall), and k is the number of effect sizes (lnRR). Positive effects indicate that the outcome measure is higher on the experimental diet (with higher BCAAs) than the control diet. B lnRR for total plasma BCAAs as a function of the BCAA content of the control diet. Red line indicates the fitted values from meta-regression, with the shaded area the CI. C–E Surfaces showing meta-regression estimates of lnRR for plasma levels of individual BCAAs as a function of the difference in dietary levels between experiment and control, and overall control levels of the focal amino acid. On surfaces, red colours indicate positive effects, blue colours negative effects and purple areas on inlaid panels indicate that the CI for that region of the surface does not span zero. All fitted values come from the AIC favoured-model (see Additional File 1: Table S5 for coefficients). F Orchard plots showing effects for each outcome stratified by whether animals were fasted or fed prior to sampling (the significance of between group contrasts are given in Additional File 1: Table S3). G Where shown, individual effect sizes are scaled by their precision