How to Write a Systematic Review: PRISMA Guide

How to Write a Systematic Review: PRISMA Guide

Learning how to write a systematic review is one of the most valuable skills in evidence-based research. Systematic reviews sit at the top of the evidence hierarchy, and following the PRISMA guidelines ensures your review is transparent, reproducible, and methodologically sound. This systematic review guide takes you through every stage of the process — from developing a protocol and registering with PROSPERO to creating the PRISMA flow diagram and avoiding common pitfalls. Whether you are conducting your first systematic review or looking to improve your methodology, this guide provides the detailed, practical instruction you need.

What Is a Systematic Review?

A systematic review is a type of research study that uses explicit, pre-defined methods to identify, select, critically appraise, and synthesize all relevant evidence addressing a specific research question. Unlike narrative reviews, which are subject to the author's selection bias, systematic reviews aim to minimize bias through methodological rigor and transparency.

The defining characteristics of a systematic review include:

• A clearly formulated research question
• A predefined and documented protocol
• A comprehensive and systematic search of multiple databases
• Explicit inclusion and exclusion criteria applied consistently
• Quality assessment (risk of bias evaluation) of included studies
• Systematic data extraction using standardized forms
• Transparent reporting of methods and findings
• Often (but not always) a quantitative synthesis (meta-analysis)

Systematic reviews are essential for evidence-based clinical decision-making, guideline development, and health policy. They are also among the most cited types of publications, making them valuable for academic careers.

PRISMA: The Reporting Standard

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) is the internationally recognized reporting guideline for systematic reviews. Originally published in 2009 and updated in 2020 (PRISMA 2020), the guideline provides a 27-item checklist and a flow diagram template that ensure transparent and complete reporting.

PRISMA 2020 introduced several updates, including:

• New items on registration, protocol, and deviations from protocol
• Updated flow diagram reflecting modern search workflows (including registers and other sources)
• Items on certainty assessment (e.g., GRADE approach)
• Abstract checklist for structured abstracts
• Emphasis on reporting not just what was done but why

Following PRISMA does not guarantee a high-quality review — it guarantees transparent reporting. The quality of your review depends on the rigor of your methods, which PRISMA helps you document clearly.

Step 1: Develop a Protocol

Every systematic review should begin with a protocol — a detailed plan that describes your research question, objectives, and methods before you start the review. The protocol serves as a roadmap and protects against outcome reporting bias (selectively reporting favorable results).

Your protocol should include:

Research question: Clearly stated using PICO (Population, Intervention, Comparison, Outcome) or an equivalent framework. Specify the primary and any secondary outcomes.

Eligibility criteria: Detailed inclusion and exclusion criteria for study design, population, intervention, comparators, outcomes, timeframe, language, and publication status.

Information sources: List all databases you will search, along with any grey literature sources, trial registries, and reference list checking strategies.

Search strategy: The full electronic search strategy for at least one database, developed with input from a librarian or information specialist if possible.

Selection process: How studies will be screened (number of reviewers, resolution of disagreements, software used).

Data extraction: What data will be extracted and how (standardized forms, piloting).

Risk of bias assessment: Which tool(s) will be used (e.g., Cochrane RoB 2 for RCTs, Newcastle-Ottawa Scale for observational studies).

Synthesis methods: How data will be synthesized (narrative, meta-analysis), including statistical methods for meta-analysis if applicable (effect measures, heterogeneity assessment, subgroup analyses, sensitivity analyses).

Step 2: Register Your Protocol with PROSPERO

PROSPERO is an international prospective register of systematic review protocols, maintained by the Centre for Reviews and Dissemination at the University of York. Registration is free and provides a public record of your review intentions, helping to prevent duplication and increase transparency.

Registration should occur before you begin data extraction (ideally before you begin screening). To register, you submit your protocol information through the PROSPERO website. After review by the PROSPERO team, your protocol is published with a unique registration number that you should cite in your final publication.

While registration is strongly encouraged, it is not universally required. However, many journals now require PROSPERO registration for systematic review submissions. Even if not required, registration demonstrates methodological rigor and good research practice.

For your literature review, also read our complete literature review guide.

Step 3: Develop a Comprehensive Search Strategy

The search strategy is arguably the most critical component of a systematic review. A poorly designed search will miss relevant studies, undermining the validity of your entire review.

Collaborate with a librarian. Information specialists have expertise in database-specific search syntax, controlled vocabularies, and search optimization. Many academic institutions provide librarian support for systematic reviews.

Search multiple databases. At minimum, search MEDLINE (via PubMed), Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL). Add discipline-specific databases as appropriate (CINAHL for nursing, PsycINFO for psychology, etc.).

Use both controlled vocabulary and free text. Controlled vocabulary (MeSH in PubMed, Emtree in Embase) captures articles tagged with standardized terms. Free-text searching captures articles not yet indexed or using non-standard terminology. Combining both approaches maximizes sensitivity.

Search structure example:

Concept 1 (Population): MeSH terms OR free-text synonyms AND Concept 2 (Intervention): MeSH terms OR free-text synonyms AND Concept 3 (Comparison — optional): MeSH terms OR free-text synonyms AND Concept 4 (Study design filter — if applicable)

Supplementary searches: In addition to database searching, conduct supplementary searches including: - Checking reference lists of included studies and relevant reviews - Searching trial registries (ClinicalTrials.gov, WHO ICTRP) for completed but unpublished studies - Searching grey literature sources (conference abstracts, thesis databases, government reports) - Contacting experts in the field for unpublished data - Forward citation searching (finding articles that cite key included studies)

Document everything. Record the complete search strategy for each database, including the date of the search and the number of results. This information is required by PRISMA and is essential for reproducibility.

Step 4: Screen Studies

Screening is the process of applying your inclusion and exclusion criteria to the studies identified by your search. It typically occurs in two stages:

Stage 1 — Title and Abstract Screening: Import all search results into reference management or systematic review software (e.g., Covidence, Rayyan, EPPI-Reviewer). Remove duplicates. Two independent reviewers screen titles and abstracts, classifying each as "include," "exclude," or "uncertain." Uncertainties are resolved through discussion or by a third reviewer. At this stage, err on the side of inclusion.

Stage 2 — Full-Text Screening: Obtain the full text of all articles that passed Stage 1. Two independent reviewers assess each article against the full eligibility criteria. Record the specific reason for excluding each article (this information is required for the PRISMA flow diagram).

Pilot screening. Before beginning formal screening, pilot your criteria on a sample of 20-50 articles. This helps ensure that reviewers interpret the criteria consistently and identifies any ambiguities.

Inter-rater reliability. Calculate a measure of agreement between reviewers (e.g., Cohen's kappa) to demonstrate consistency. A kappa value above 0.7 is generally considered acceptable.

Step 5: Extract Data

Create a standardized data extraction form and pilot it on 3-5 included studies before applying it to all studies. Data extraction should be performed independently by two reviewers, with discrepancies resolved through discussion.

Your extraction form should capture:

• Study identifiers (first author, year, country, journal, funding source)
• Study design and duration
• Participant characteristics (sample size, age, sex, clinical condition, severity)
• Intervention details (type, dose, duration, frequency, mode of delivery)
• Comparator details
• Outcome definitions and measurement methods
• Results (effect estimates, confidence intervals, p-values, sample sizes per group)
• Risk of bias assessment items
• Additional notes

For meta-analysis, you may need to contact study authors for missing data or to clarify ambiguous reporting.

Step 6: Assess Risk of Bias

Risk of bias assessment evaluates the internal validity of each included study. Different tools are used depending on the study design:

Randomized Controlled Trials: Use the Cochrane Risk of Bias tool (RoB 2), which assesses five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each domain is judged as "low risk," "some concerns," or "high risk."

Non-Randomized Studies of Interventions: Use ROBINS-I (Risk of Bias in Non-Randomized Studies of Interventions), which assesses seven domains including confounding, selection, and measurement bias.

Observational Studies: The Newcastle-Ottawa Scale (NOS) is commonly used for cohort and case-control studies, assessing selection, comparability, and exposure/outcome assessment.

Present risk of bias results in summary tables or traffic-light plots. Discuss how the risk of bias findings affect your confidence in the overall evidence.

Step 7: Assess Certainty of Evidence (GRADE)

The GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) approach provides a framework for rating the certainty of evidence across studies for each outcome. GRADE considers five domains:

1. **Risk of bias:** Overall methodological quality of included studies
2. **Inconsistency:** Unexplained heterogeneity in results across studies
3. **Indirectness:** Differences between the review question and the available evidence
4. **Imprecision:** Wide confidence intervals or small sample sizes
5. **Publication bias:** Evidence of missing studies (assessed using funnel plots, Egger's test)

Evidence is rated as high, moderate, low, or very low certainty. GRADE Summary of Findings tables are increasingly required by journals and are a core component of Cochrane reviews.

Step 8: Synthesize Results

Narrative synthesis organizes and describes findings across studies, identifying patterns, explaining differences, and drawing conclusions. This is appropriate when studies are too heterogeneous for statistical combination.

Meta-analysis quantitatively combines results using statistical methods. Key considerations include:

• **Effect measures:** Risk ratios or odds ratios for dichotomous outcomes; mean differences or standardized mean differences for continuous outcomes
• **Model selection:** Fixed-effect models assume a common true effect; random-effects models allow for between-study variation. Random effects are more conservative and generally more appropriate when clinical or methodological heterogeneity is expected.
• **Heterogeneity assessment:** Use the I-squared statistic (percentage of variation due to heterogeneity rather than chance) and the chi-squared test. I-squared values above 50% suggest substantial heterogeneity.
• **Subgroup and sensitivity analyses:** Pre-planned analyses that explore potential sources of heterogeneity or test the robustness of results (e.g., excluding high-risk-of-bias studies).
• **Publication bias:** Assessed using funnel plots (visual inspection for asymmetry) and statistical tests (Egger's test, Begg's test).

Step 9: Create the PRISMA Flow Diagram

The PRISMA flow diagram visually summarizes the study selection process. The PRISMA 2020 flow diagram includes:

• Number of records identified from each database and other sources
• Number of duplicates removed
• Number of records screened (titles and abstracts) and excluded
• Number of reports sought for retrieval and not retrieved
• Number of reports assessed for full-text eligibility and excluded (with reasons)
• Number of studies included in the review (and in meta-analysis, if applicable)

The flow diagram should be included as a figure in your review. Templates are available on the PRISMA website (prisma-statement.org).

Step 10: Complete the PRISMA Checklist

Before submitting your review for publication, complete the PRISMA 2020 checklist. The checklist has 27 items covering every section of the manuscript (title, abstract, introduction, methods, results, discussion, and other information). For each item, indicate the page number or section where the information can be found.

Many journals require the PRISMA checklist to be submitted alongside the manuscript. Even when not required, completing the checklist ensures you have not omitted important reporting elements.

Common Pitfalls to Avoid

1. **Starting without a protocol.** Ad hoc decisions during the review process introduce bias. Always develop and register a protocol before starting.

1. **Insufficient search strategy.** Searching only one database or using overly narrow search terms compromises comprehensiveness. Collaborate with a librarian and search multiple sources.

1. **Single-reviewer screening and extraction.** Dual independent screening and extraction is the standard for minimizing errors and bias. Single-reviewer approaches should be acknowledged as a limitation.

1. **Ignoring grey literature.** Publication bias (the tendency for studies with positive results to be published) can distort review findings. Searching grey literature helps mitigate this.

1. **Not assessing risk of bias.** Including studies without evaluating their quality is a significant methodological flaw. Always use validated assessment tools.

1. **Inappropriate meta-analysis.** Combining studies that are too heterogeneous in design, population, or outcomes produces misleading pooled estimates. Meta-analysis should only be performed when studies are sufficiently similar.

1. **Incomplete reporting.** Failing to report all PRISMA items, omitting the flow diagram, or not providing the full search strategy reduces the transparency and reproducibility of your review.

1. **Not updating the search.** If there is a long gap between your initial search and manuscript submission, update your search to capture recently published studies.

Conclusion

Writing a systematic review is a rigorous, structured process that requires careful planning, methodological discipline, and transparent reporting. By following the steps outlined in this PRISMA guide — from protocol development to final reporting — you can produce a review that meaningfully contributes to evidence-based practice and academic knowledge. The investment in learning these methods pays dividends throughout your research career, as systematic review skills are applicable to grant applications, clinical guideline development, and evidence synthesis in any field.

For related guidance, explore our meta-analysis guide and our comprehensive literature review guide.

---

**Simplify your systematic review workflow.** PubMEDIS provides AI-powered search, screening assistance, and evidence synthesis tools that help you complete systematic reviews faster and more accurately. Create your free account and start your review today.