Skip to main content
. 2019 Sep 12;52(11):1205–1223. doi: 10.1002/eat.23158

Figure 2.

Figure 2

Summary plot of all 88 meta‐analyses comparing female AN cases with healthy controls. The plot shows the Q values (i.e., the false discovery rate‐corrected p values) of each inverse variance‐weighted random‐effects meta‐analyses comparing AN cases pretreatment (purple, n = up to 2,294), posttreatment (light green, n = up to 722), and after weight recovery (dark green, n = up to 398) with healthy controls (n = up to 2,251). Restricted maximum‐likelihood estimator was used to estimate heterogeneity. Q values are transformed on the ‐log10 scale, multiplied by the sign of the mean difference (MD) and presented on the x‐axis. Points lying in the yellow area indicate no statistically significant mean difference between AN cases and healthy controls after correction for multiple testing (i.e., Q > 0.05). Points to the left of the yellow area indicate a lower mean value in AN cases than in controls, whereas points on the right of the yellow area indicate a higher mean value in AN cases than in controls. A dark green point outside the yellow area indicates a significant difference between AN cases and controls after weight recovery (a) The outcomes with the largest differences between cases and controls. (b) The less extreme mean differences. The x‐axis was capped at ‐log10(Q value) × sign(MD) = −0.25. The full figure is presented as Figure S3