Read a meta-analysis by tracing eligibility, effect measures, model choices, heterogeneity, sensitivity analyses, and certainty back to the included studies.
Confirm the review question and eligibility rules
Write down the population, intervention or exposure, comparator, outcomes, study designs, dates, and language limits. The pooled result only applies to the evidence the authors chose and were able to find.
Inspect the search and study flow
Check databases, search dates, screening process, exclusions, and the study-flow diagram. Missing searches or opaque exclusions can shape the result before any statistical model is run.
Understand the effect measure
Determine whether the plot uses a risk ratio, odds ratio, mean difference, standardized mean difference, hazard ratio, or another measure. Note direction, units, confidence interval, and the threshold for practical—not only statistical—importance.
Read heterogeneity as a research question
A single I-squared value is not an explanation. Compare populations, interventions, follow-up, measurement, design quality, and context. Subgroup results should be interpreted cautiously, especially when not prespecified.
Check robustness and certainty
Review risk-of-bias judgments, sensitivity analyses, publication-bias assessments, influence of large studies, and the authors’ certainty framework. A precise pooled number can still rest on indirect or biased evidence.
Frequently asked questions
What does the diamond in a forest plot mean?
It usually represents the pooled effect and its confidence interval; confirm the plot legend and effect direction.
Is high I-squared always fatal?
No. It signals inconsistency to investigate, not an automatic verdict.
Does statistical significance mean practical importance?
No. Interpret effect size, interval, baseline risk, harms, costs, and context.
Can a meta-analysis fix weak studies?
No. Pooling does not remove bias or poor measurement in the included evidence.
What is a sensitivity analysis?
It tests whether conclusions change under different inclusion, model, or analytic choices.
Should I read the included studies?
Read the key studies and those driving the result when the decision matters.
Reconstruct the pooled claim
Pick the headline outcome and write a one-sentence version that includes population, intervention or exposure, comparator, time horizon, effect measure, and uncertainty interval. Then identify which studies contribute most weight and whether their settings resemble the decision you care about. Check whether sensitivity analyses, subgroup choices, or excluded studies materially change the result. If certainty is low or heterogeneity is substantial, keep that limitation in the sentence rather than burying it in a note. The goal is not to recalculate the review; it is to understand exactly what was pooled and how far the pooled estimate can reasonably travel.
CTA: Start reading with Readever and build a source-linked understanding you can audit later.
Related Readever pages
- How to read scientific papers
- Research paper summary template
- How to annotate a research paper PDF
- Literature review matrix template
