What is the target effect?

Fixed-effect models assume one true effect shared by all studies.
Random-effects models assume each study has its own true effect.
Mixed-effects models allow fixed predictors while accounting for random variation.

Fixed-effect: when simplicity misleads

This model ignores between-study heterogeneity completely.
It often produces narrow confidence intervals that look convincing.
Useful only when studies are nearly identical in design and population.

Random-effects: not a universal solution

Random-effects account for heterogeneity but change study weighting.
Smaller studies gain more influence, which may increase bias.
The pooled estimate represents an average that may fit no single population.

Mixed-effects: explaining heterogeneity

Mixed-effects models include study-level variables as fixed effects.
They help identify why results differ across studies.
Despite their power, they are underused due to complexity.

Why model choice changes conclusions

Different models can produce different effect sizes and certainty.
This directly affects clinical interpretation and guideline development.
Choosing a model without justification risks misleading conclusions.

Choosing wisely, not habitually

I² alone should not dictate model selection.
Clinical diversity and study design matter just as much.
Transparent reporting of model choice strengthens meta-analysis credibility.