The b coefficient in the meta-regression with BMI variations treated as a covariate was 0.008795, with a 95% CI
of 0.0248 to 0.00675, which is narrow and includes 0, suggesting that mean BMI does not explain variations in
treatment effects among studies and across treatment comparisons. Sensitivity analysis on the basis of fixed-effect
methods confirmed the results found with the randomeffects meta-analysis, as listed in Table 3.