How do I fix a meta-analysis that pooled studies with incompatible designs?

Reclassify every included study by design, then separate incompatible designs into distinct analyses or switch to a narrative synthesis, and document how that decision changes the conclusions.

A pooled estimate assumes the studies are similar enough that combining their effect measures answers one question. When designs differ, that assumption needs an explicit justification. The Cochrane Handbook's meta-analysis guidance covers effect measures, fixed and random effects, heterogeneity, sensitivity analyses, and careful interpretation of pooled results, which are the elements you re-examine once designs are separated.

1. Build a design table for every included study. Record the design label, the effect measure reported, and the comparison being made. The result is a visible list of which studies share a design and outcome scale.
2. Mark each pooled analysis as compatible or mixed. A group is compatible when its studies share a design and report an effect measure that can be combined. Flag any group that mixes designs without a stated rationale.
3. Re-run the compatible groups as separate analyses. Report each design-specific estimate with its confidence interval and heterogeneity, following the Cochrane guidance on effect measures and forest plots. The result is a set of estimates you can interpret on their own terms.
4. Use a narrative synthesis for studies that cannot be pooled. Describe their findings and direction without a combined estimate, and state why pooling was set aside.
5. Run a sensitivity analysis that removes the design-based grouping choice. Compare the separated results with the original pooled result. A large shift tells you the pooling decision drove the conclusion.
6. Rewrite the conclusions to match the separated evidence. Cochrane's interpretation guidance directs attention to confidence intervals, absolute and relative effects, applicability, and certainty. State which conclusions rest on pooled estimates and which rest on narrative description.
7. Document the decision in the methods and results. Record the design classification, the analyses separated, and the effect on conclusions so reviewers can trace the change.

Design-specific quality assessment supports this step. The NHLBI Study Quality Assessment Tools are organized by design and focus reviewers on internal validity and methodological flaws, with users defining their own judgement parameters. Applying the matching tool to each design group keeps the appraisal aligned with the analysis.