Table 3.

Fixed effect model versus random effect model

	Fixed effects model	Random effects model
Treatment effect	Assumed that the treatment effect of one agent compared with another is the same across all trials containing those treatments	Assumed that the treatment effect size varies between studies
Study weights	Based upon within-study variance only	Based upon within-study and between-study variances
Use	If there are a small number of studies with large sample sizes (i.e., sparse evidence base) If there are too few studies to accurately estimate between-study variance	If substantial heterogeneity in study or patient characteristics between the combined studies is suspected

Source [64]