Systematic review with pairwise meta-analysis
What is the effect of one intervention versus a specified comparator?
When to use this project
Use this starter for a focused systematic review when studies provide sufficiently compatible estimates for one or more pairwise comparisons. Define population, intervention or exposure, comparator, outcome and follow-up before choosing a model.
How it differs
Pairwise synthesis combines direct comparisons. A network review connects several treatment comparisons; a scoping review maps a field. Not every outcome in this project has to be pooled.
Decisions before you start
- Choose outcome-specific effect measures: risk ratio or odds ratio for binary data, mean difference for a shared scale, standardized mean difference for compatible constructs measured on different scales, and an appropriate ratio for rates or hazards. These measures are not interchangeable.
- Prespecify how multi-arm studies, repeated time points, cluster designs and adjusted estimates contribute without double counting.
- Explain the pooling decision clinically and methodologically. A heterogeneity statistic alone cannot decide whether pooling is appropriate.
How the flows work together
Each flow below links to its diagram, steps and tables. Flow order organizes the project; it does not imply that all work is strictly sequential. Data assessment, analysis planning and protocol refinement may inform one another.
1. Review overview
Overview · Question through evidence interpretation
Use this overview to state the purpose and link to detailed decisions. Keep a versioned protocol and decision register, and add evidence references for the rationale.
2. Search and study selection
Selection · Identify and screen eligible studies
Keep records, reports and studies as separate counting units. The starter partition covers duplicate removal only: add explicit branches for automation or other removals if used. Supplementary searches may need separate identification routes. A final report-to-study mapping is not an exclusion count.
3. Data extraction and synthesis
Statistics · Prepare and combine study estimates
Use the alternative synthesis path when pooling is inappropriate. Complete effect-scale, dependence, heterogeneity and uncertainty specifications before filling a result shell. Weighted-mean notation is not an implemented calculation.
4. Bias, robustness and certainty
Statistics · Assess trustworthiness of findings
Study risk of bias, sensitivity analysis and outcome-level certainty are distinct activities. Select appropriate tools and state their versions rather than assigning an unexplained quality score.
Equations and statistical interpretation
These explanations describe notation and assumptions. Flow does not calculate these values.
Inverse-variance random-effects mean
yᵢ is the study effect on its analysis scale; sᵢ² is its sampling variance; τ² is between-study variance. Ratio effects are commonly analysed on the log scale.
This is model notation, not a complete estimator specification. State the variance estimator and interval method; sparse binary data may require a different approach.
Published example
Sherrington et al. (2019), Exercise for preventing falls in older people living in the community
A published intervention review illustrating pairwise synthesis. Inspect how fall rates and people experiencing a fall are treated as different outcomes.
This publication illustrates a study design, not an endorsement of every method or a reproduction of its results. Follow its protocol, supplement and update history when studying the example.
Hypothetical scenario
A rehabilitation team wants to estimate whether balance exercise reduces falls compared with usual care among community-dwelling older adults.
- Review overview: Specify the population, exercise definition, usual-care comparator and follow-up in the overview; record the protocol and any registration.
- Search and study selection: Search for eligible trials, link multiple reports from the same trial, and record full-text exclusions.
- Data extraction and synthesis: Extract fall counts and person-time for a rate-ratio analysis; keep the proportion of participants who fall as a separate binary outcome.
- Bias, robustness and certainty: Compare justified sensitivity analyses and assess certainty separately for each outcome before writing conclusions.
Deliver a populated protocol, decision log, data specification, and planned outputs. Enter numerical findings only after analysis and verification. Keep decisions made before seeing results distinguishable from later amendments.
Sources
- Cochrane Handbook, Chapter 10: Meta-analysis
- PRISMA 2020 reporting guidance
- PRISMA 2020 flow diagrams: identification and selection
- Cochrane Handbook, Chapter 14: Certainty and Summary of findings
Source check: 2026-09-17. Record the version of each reporting or appraisal tool used in your actual protocol.
Compare all Evidence synthesis project types · Open editor in a new tab