How to Design a Systematic Review
To design a high-quality systematic review, follow the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, which provide a structured approach with explicit, reproducible methods to minimize bias and ensure comprehensive evidence synthesis. 1
Step 1: Define Your Research Question
- Use the PICOS framework to structure your question:
- Population: Define the specific group of participants
- Intervention/exposure: Clearly specify the intervention being studied
- Comparator: Identify what the intervention is being compared with
- Outcomes: Specify primary and secondary outcomes of interest
- Study designs: Determine which study types will be included 1
Example: "In patients with chronic renal failure (P), does hepatitis B vaccination (I) compared to no vaccination (C) improve seroconversion rates (O) based on randomized controlled trials (S)?"
Step 2: Develop a Protocol
- Register your protocol (e.g., on PROSPERO) to enhance transparency
- Include in your protocol:
- Background and rationale
- Explicit research question
- Inclusion/exclusion criteria
- Search strategy
- Study selection process
- Data extraction methods
- Risk of bias assessment approach
- Data synthesis plan 1
Step 3: Conduct a Comprehensive Search
- Search multiple databases (MEDLINE alone is insufficient)
- Include:
- Document your search strategy completely so it can be reproduced
- Work with a librarian or information specialist if possible
Step 4: Study Selection
- Screen titles and abstracts against inclusion/exclusion criteria
- Review full texts of potentially eligible studies
- Have two independent reviewers perform selection with a process to resolve disagreements
- Document the selection process using a PRISMA flow diagram 1, 4
- Record reasons for exclusion of full-text articles
Step 5: Data Extraction
- Create standardized data extraction forms
- Extract data in duplicate to minimize errors
- Collect information on:
Step 6: Risk of Bias Assessment
- Use appropriate tools based on study design:
- Cochrane Risk of Bias tool for randomized trials
- ROBINS-I for non-randomized intervention studies
- QUADAS-2 for diagnostic accuracy studies
- Have two independent reviewers perform assessments
- Consider how bias might affect the interpretation of results 1, 3
Step 7: Data Synthesis and Analysis
- Decide between narrative synthesis or meta-analysis based on study heterogeneity
- For meta-analysis:
- Choose appropriate effect measures (risk ratio, odds ratio, mean difference)
- Select fixed or random effects models
- Assess heterogeneity (I² statistic)
- Conduct sensitivity analyses
- Consider subgroup analyses (if pre-specified in protocol)
- Evaluate certainty of evidence using GRADE approach 1, 5
Step 8: Reporting Results
- Follow the PRISMA checklist for reporting
- Include:
- Structured abstract
- Clear description of methods
- PRISMA flow diagram
- Summary of included studies
- Risk of bias assessments
- Results of syntheses
- Discussion of limitations
- Implications for practice and research 1
Common Pitfalls to Avoid
- Poorly defined research question
- Inadequate search strategy (using only one database)
- Selection bias (not having independent duplicate screening)
- Data extraction errors (not using standardized forms)
- Inappropriate synthesis methods (forcing meta-analysis when heterogeneity is high)
- Failure to assess publication bias
- Drawing conclusions beyond what the evidence supports 6, 3
Tools to Facilitate the Process
- Protocol registration: PROSPERO
- Literature management: Covidence, Rayyan, EndNote
- Risk of bias assessment: RevMan, robvis
- Meta-analysis: RevMan, Comprehensive Meta-Analysis, R packages (metafor, meta)
- Reporting: PRISMA checklist and flow diagram 6, 4
By following these steps and adhering to the PRISMA guidelines, you can design and conduct a systematic review that provides reliable evidence synthesis to inform clinical practice and future research.