To write a systematic review, follow nine sequential steps: formulate a PICO research question, assemble a multidisciplinary team, develop and register a protocol on PROSPERO, execute a comprehensive multi-database search, conduct dual-reviewer screening, extract data into structured matrices, assess risk of bias using validated tools, synthesize findings narratively or via meta-analysis, and write the PRISMA 2020-compliant manuscript. Each step builds on the last and cannot be reversed without a documented protocol amendment.
The following table indicates the 9 steps for writing a systematic review.
Table 1: How to write a systematic review in 9 steps
| # | Step | Key actions |
| 1 | Formulate your research question PICO | Define Population, Intervention, Comparison, and Outcome. Use SPIDER for qualitative questions or PECO for epidemiological studies. Confirm no equivalent review exists on PROSPERO or Cochrane Database. |
| 2 | Assemble your review team | Minimum four specialists: principal investigator, second independent reviewer, medical librarian or information specialist, and biostatistician. Teams below three members fail dual-reviewer requirements. |
| 3 | Write and register your protocol PROSPERO | Document eligibility criteria, databases, search terms, risk-of-bias tools, and synthesis method using the PRISMA-P 17-item checklist. Register on PROSPERO before any searching begins. Registration takes 5 to 10 business days. |
| 4 | Design and execute the literature search Boolean | Build search strings using MeSH terms and free-text synonyms connected by AND between PICO elements and OR within them. Search at minimum PubMed/MEDLINE, Embase, and Cochrane CENTRAL. Document the exact string, date, and record count per database. |
| 5 | Screen studies for eligibility | Two independent reviewers screen title/abstract then full-text records. Use Covidence or Rayyan for workflow management. AI-assisted screening reduces workload by 50 to 70%. Calculate Cohen's kappa for inter-rater agreement. |
| 6 | Extract data from included studies | Use a pre-piloted extraction form capturing author, year, design, sample size, intervention, comparator, outcomes, effect estimates, and funding source. Two reviewers extract independently and cross-check. Platforms: Covidence, Excel, SysRev. |
| 7 | Assess risk of bias Appraisal | Apply RoB 2 for RCTs, ROBINS-I for observational studies, QUADAS-2 for diagnostic accuracy studies, JBI Checklist for qualitative studies, and AMSTAR-2 for umbrella reviews. Summarize in a traffic-light plot. Apply GRADE to rate certainty of evidence. |
| 8 | Synthesize the evidence | Choose narrative synthesis when studies are too heterogeneous to pool. Run meta-analysis when two or more studies report the same outcome with compatible metrics. Outputs: pooled effect estimate, forest plot, I-squared, Cochran Q, and funnel plot. Platforms: RevMan 5, R (meta, metafor), Stata. |
| 9 | Write the manuscript PRISMA 2020 | Structure in IMRAD format. Apply all 27 PRISMA 2020 checklist items. Include structured abstract, full search strategy appendix, PRISMA flow diagram, characteristics table, risk-of-bias plots, forest plots, and GRADE evidence table. |
Step 1: Formulate Your Research Question Using PICO
To formulate a systematic review research question, use the PICO framework: Population, Intervention, Comparison, and Outcome. PICO is the standard question structure recommended by the Cochrane Handbook for intervention reviews. Alternate frameworks include SPIDER for qualitative questions and PECO for epidemiological exposure-outcome questions.
A well-formed PICO question is narrow, answerable, and not previously answered by an existing high-quality review. Before writing a protocol, search PROSPERO and the Cochrane Database of Systematic Reviews to confirm no equivalent review has been completed in the last three years. Example: ‘In postmenopausal women with osteoporosis (P), does calcium and vitamin D supplementation (I) compared to placebo (C) reduce hip fracture incidence (O)?’
Step 2: Assemble Your Review Team
A systematic review team requires a minimum of four specialists. The University of California Irvine Library specifies a principal investigator, a second independent reviewer, a medical librarian or information specialist, and a statistician. Teams with fewer than three members produce reviews that journals reject for failing dual-reviewer requirements.
1. Principal investigator (PI) - Subject-matter expert who formulates the research question, approves the protocol, interprets findings, and takes authorship responsibility under ICMJE criteria.
2. Second reviewer - Independent screener who screens all records and extracts data separately. Disagreements are documented and resolved by a third arbiter.
3. Medical librarian or information specialist - Designs and peer-reviews the search strategy using the PRESS checklist. Searches at least three major databases.
4. Biostatistician - Performs meta-analysis, calculates effect sizes, generates forest plots, assesses heterogeneity (I-squared, Cochran Q), and runs sensitivity analyses.
Step 3: Write and Register Your Protocol
To write a systematic review protocol, document the research question, eligibility criteria, planned databases, search terms, screening procedures, data extraction variables, risk-of-bias tools, and synthesis method. The PRISMA-P (Protocol) 17-item checklist governs protocol reporting.
Submit the completed protocol to PROSPERO before executing any database searches. PROSPERO registration takes 5 to 10 business days. Once registered, include the PROSPERO ID in the manuscript's Methods section. Protocol amendments made after registration must be documented in PROSPERO with a date and rationale. Conducting searches before registration constitutes a reporting violation under PRISMA 2020 Item 24.
Step 4: Design and Execute the Literature Search
To design a systematic review search strategy, construct Boolean search strings using MeSH terms and free-text synonyms for each PICO element, connected by AND between elements and OR within elements. The minimum database set recommended by AHRQ, CRD, and Cochrane is PubMed/MEDLINE, Embase, and Cochrane CENTRAL.
A 2017 prospective study published in Systematic Reviews found that the combination of Embase, MEDLINE, Web of Science, and Google Scholar achieved 98.3% recall across systematic reviews. Adding CINAHL for allied health or PsycINFO for mental health increases topic-specific recall. Document the exact search string, date, and record count per database for the PRISMA flow diagram.
Step 5: Screen Studies for Eligibility
Screening proceeds in two sequential stages. Title and abstract screening removes clearly ineligible records. Full-text screening applies the pre-specified eligibility criteria to remaining records. Both stages require two independent reviewers without knowledge of each other's decisions.
Covidence is endorsed by Cochrane as its standard review management platform. Rayyan is free and uses AI-assisted relevance prediction, reducing screening time by 50 to 70%. Screening 5,000 records at the title/abstract stage requires approximately 120 person-hours. Calculate Cohen's kappa to measure inter-rater agreement and report it in the Methods section.
Step 6: Extract Data from Included Studies
To extract data from included studies, use a pre-piloted extraction form capturing author, year, country, study design, sample size, participant characteristics, intervention details, comparator, outcome measures, effect estimates, follow-up period, and funding source. Two reviewers extract data independently and cross-check entries.
Data extraction tables form a core component of the published manuscript's supplementary materials. Journals, including the Annals of Internal Medicine, require complete extraction tables at submission. Software platforms for data extraction include Covidence, Excel, and SysRev (open-source, free).
Step 7: Assess Risk of Bias
Risk-of-bias assessment evaluates the methodological validity of each included study. The tool used depends on the study design.
| Study Design | Risk-of-Bias Tool | Domains Assessed |
| Randomized controlled trial | Cochrane RoB 2 | Randomization, allocation concealment, blinding, outcome reporting, other bias |
| Non-randomized observational study | ROBINS-I | Confounding, selection, classification, deviations, missing data, measurement, reporting |
| Diagnostic accuracy study | QUADAS-2 | Patient selection, index test, reference standard, flow and timing |
| Qualitative study | JBI Critical Appraisal Checklist | Research design, data collection, analysis, interpretation |
| Umbrella review | AMSTAR-2 | Protocol registration, search, study selection, data extraction, bias, synthesis |
Risk-of-bias judgments are summarized in a traffic-light plot. GRADE then classifies overall certainty of evidence for each outcome as high, moderate, low, or very low. GRADE tables are required by journals including JAMA and Annals of Internal Medicine.
Step 8: Synthesize the Evidence
To synthesize evidence in a systematic review, choose between narrative synthesis and meta-analysis based on the clinical and statistical homogeneity of included studies. Narrative synthesis organizes findings thematically when studies are too heterogeneous to pool. Meta-analysis statistically pools effect estimates from two or more studies reporting the same outcome with compatible metrics.
Meta-analysis outputs include a pooled effect estimate with a 95% CI, a forest plot, an I-squared statistic (values above 50% indicate substantial heterogeneity), a Cochran Q p-value, and a funnel plot with an Egger test for publication bias. Statistical platforms are RevMan 5, R packages meta and metafor, and the Stata metan command.
Step 9: Write the Systematic Review Manuscript
To write a systematic review manuscript, structure it in IMRAD format: Introduction, Methods, Results, and Discussion, with a structured abstract and reference list. Apply the PRISMA 2020 27-item checklist to verify every required reporting element is present.
1. Title - Must identify the document as a systematic review or meta-analysis and include the population and outcome. Example: 'Aerobic Exercise and Glycemic Control in Adults with Type 2 Diabetes: A Systematic Review and Meta-Analysis.'
2. Abstract - Structured abstract with background, objectives, data sources, eligibility criteria, synthesis methods, results, limitations, and conclusions.
3. Introduction - Background establishing why the review is needed, the specific research question in PICO format, and the review objectives.
4. Methods - Full search strategy for at least one database in an appendix, eligibility criteria, screening procedures, data extraction process, risk-of-bias tools, synthesis method, and PROSPERO registration number.
5. Results - PRISMA flow diagram, characteristics of included studies table, risk-of-bias summary, synthesis results with forest plots or narrative tables, and GRADE evidence table.
6. Discussion - Interpretation of main findings, limitations (search restrictions, publication bias, heterogeneity), implications for practice and policy, and future research recommendations.
Conclusion
Writing a systematic review requires a structured, team-based approach across nine sequential steps, from formulating a PICO research question through delivering a PRISMA 2020-compliant manuscript. Each step builds directly on the last. Skipping protocol registration on PROSPERO, conducting single-reviewer screening, or omitting formal risk-of-bias assessment produces a review that peer-reviewed journals reject on methodological grounds.
The process demands a minimum of four specialists, access to at least three major databases, and 6 to 18 months of dedicated effort for an independent research team. Researchers who need a publication-ready systematic review without the full time investment work with a professional systematic review writing service that provides dual-reviewer screening, PROSPERO registration, biostatistician-led meta-analysis, and PRISMA-compliant manuscript delivery from protocol to submission.
To understand what a systematic review is before beginning the writing process, read the complete guide on what is a systematic review. For end-to-end support across all nine steps, visit Systematic Review Writing Services to request a scope of work.