Report / evidence synthesis

Reading a systematic review: what the combined evidence can show

A pooled result is only one part of a review. The question, selection of studies, risk of bias, missing evidence and certainty of each outcome determine what the combined findings can support.

Begin with the question, not the pooled number

Identify the population, intervention, comparison and outcomes before reading the conclusion. Cochrane uses these parts of the question to describe what a review is trying to answer, including benefits and harms. A review about one population or comparison may not answer a nearby question. Record what was eligible and what was outside scope. The title and abstract help locate the question, but the methods and included studies show which evidence actually contributes to the answer.

Evidence: Cochrane Library: what a systematic review contains / Cochrane Handbook, Chapter 10: meta-analysis / Cochrane Handbook, Chapter 15: interpreting results

Distinguish a systematic review from its statistical synthesis

A systematic review brings together evidence using explicit methods for finding, selecting and assessing studies. A meta-analysis statistically combines results from two or more studies. PRISMA covers reviews with and without a synthesis, so the absence of a pooled estimate does not by itself mean a review is unsystematic. Conversely, producing a pooled number does not establish that the review is sound. Cochrane warns that combining data before checking the question, selection, study methods and risk of bias can mislead.

Evidence: Cochrane Library: what a systematic review contains / PRISMA 2020: transparent reporting of systematic reviews / Cochrane Handbook, Chapter 10: meta-analysis

Inspect the route from search to included evidence

Look for the sources searched, the search strategies, eligibility criteria and how records moved into the included set. PRISMA asks for transparent reporting of these methods and provides a flow diagram. Distinguish a study from its reports: one trial may have a protocol, a main paper, a harms report and further analyses. Check the linked protocol or registration and any reported changes rather than assuming the final report was the original plan. Clear reporting lets a reader examine decisions; it does not certify that those decisions were appropriate.

Evidence: Cochrane Library: what a systematic review contains / PRISMA 2020: transparent reporting of systematic reviews

Keep bias visible before and after pooling

Read the risk-of-bias assessment and its reasons, not only how many studies were included. Cochrane’s synthesis guidance identifies study design, within-study biases, variation between studies and reporting biases as threats to an overall estimate. More data can improve statistical precision without removing those threats. PRISMA is a reporting guideline, not a tool for rating a review’s methodological quality. A completed checklist can make methods easier to inspect while leaving important weaknesses in the evidence or review process.

Evidence: Cochrane Library: what a systematic review contains / PRISMA 2020: transparent reporting of systematic reviews / Cochrane Handbook, Chapter 10: meta-analysis

Ask whether the combined estimate hides important differences

Studies may differ in participants, interventions, comparisons, outcomes and their effect estimates. Cochrane asks reviewers to consider whether numerical results should be combined at all. In a forest plot, individual estimates and confidence intervals matter alongside the summary diamond. Random-effects methods allow for variation in underlying effects, but they still require careful interpretation. Sensitivity analyses can show whether conclusions change with important analytical decisions. A single overall result is not proof of one identical effect in every setting.

Evidence: Cochrane Handbook, Chapter 10: meta-analysis / Cochrane Handbook, Chapter 15: interpreting results

Look for evidence that could be missing

Cochrane distinguishes an unavailable study from a missing result within an identified study. Either can bias a synthesis when reporting depends on the direction, size or statistical result of the finding. Searching sources beyond journal papers, including trial registers and regulatory material, can help reveal gaps. A funnel plot may offer clues, but its limitations prevent it from being a simple pass-or-fail test for publication bias. Describe what was checked and what remains unknown rather than treating no detected warning as proof that nothing is missing.

Evidence: PRISMA 2020: transparent reporting of systematic reviews / Cochrane Handbook, Chapter 13: missing evidence

Read certainty and effect size outcome by outcome

Cochrane’s GRADE approach rates certainty for each outcome, not as one universal label for the entire review. It considers risk of bias, inconsistency, indirectness, imprecision and publication bias. A summary-of-findings table can place the estimate, its uncertainty and reasons for the rating together. Read the effect measure, comparison and time frame; where appropriate, consider absolute as well as relative effects. A wide interval or sparse evidence is not interchangeable with proof of no effect, and a narrow interval alone does not remove bias.

Evidence: Cochrane Handbook, Chapter 14: certainty and summary of findings / Cochrane Handbook, Chapter 15: interpreting results

Separate the evidence summary from a care recommendation

Cochrane asks review authors to explain findings and the balance of benefits and harms without making decisions that depend on assumptions about resources, preferences and context. Applicability includes how well the populations, interventions and outcomes match the question at hand. Local feasibility and what matters to people also affect decisions. A useful report states the supported finding, its certainty, important gaps and where the evidence may not transfer. This guide does not turn a review into personal treatment advice or claim that Mynd has clinically reviewed its conclusions.

Evidence: Cochrane Library: what a systematic review contains / Cochrane Handbook, Chapter 15: interpreting results

Source note

This is general health-science education. No clinical review has been performed.

General education for reading reviews of health interventions, not a full appraisal instrument, a complete review-conduct manual, a clinical recommendation or a Mynd evidence-review service. Diagnostic, qualitative and other review types can require additional methods. No clinical review has been performed.