Background: Moderator effect assessment is important in personalized medicine. We mathematically prove that the average summary value is actually nonlinearly to logRR, and we assess the bias from linear meta‑regression on logRR via simulation. Methods: In the meta‑analysis of randomized controlled trials, the moderator effect is generally evaluated by the linear meta‑regression of the logarithmic risk ratio (RR) versus the average summary value of the entire study population. Conclusions: We recommend using linear meta‑regression on logarithmic odds ratio (logOR) since it has been shown that the average summary value is actually linear to logOR.