How to Do a Systematic Review
A step-by-step guide to conducting a rigorous systematic literature review, from formulating your question to publishing your findings.
What Is a Systematic Review?
A systematic review is a structured, transparent method of identifying, evaluating, and synthesizing all available research on a specific question. Unlike a traditional literature review, a systematic review follows a pre-defined protocol to minimize bias and ensure reproducibility.
Protocol-Driven
Methods are defined before the review begins
Comprehensive Search
Multiple databases searched systematically
Transparent Reporting
Every decision is documented and reproducible
Define Your Research Question
Every systematic review starts with a clear, focused research question. The most effective way to structure this is the PICO framework:
Who are you studying?
What treatment/exposure?
Against what?
What are you measuring?
Example: "In adults with type 2 diabetes (P), does metformin (I) compared to lifestyle changes alone (C) reduce HbA1c levels (O)?"
Write a Protocol
Before you start searching, document your methodology in a review protocol. This prevents cherry-picking results and increases credibility.
Your protocol should include:
- ✓ Research question (PICO)
- ✓ Inclusion and exclusion criteria
- ✓ Databases to search (PubMed, Scopus, Web of Science, etc.)
- ✓ Search strategy with keywords and Boolean operators
- ✓ Screening process (single vs. dual screening)
- ✓ Data extraction plan
- ✓ Quality assessment method
Search the Literature
Search at least 2–3 databases to ensure comprehensive coverage. Common choices include PubMed, Scopus, Web of Science, and Embase.
Search Strategy Tips
- • Use Boolean operators (AND, OR, NOT)
- • Include MeSH terms and free-text keywords
- • Use truncation (e.g., "therap*" for therapy, therapies)
- • Apply date and language filters only when justified
- • Save your search history for reproducibility
⚡ Speed Up with Lumina
Lumina lets you search seven supplementary scholarly and government sources in parallel. Government queries can expand across five languages, and the resulting grey literature remains clearly labelled before import.
Start your 14-day trial →Screen Titles & Abstracts
This is typically the most time-consuming step. You'll review each paper's title and abstract against your inclusion criteria and decide whether to include, exclude, or mark it as maybe.
Best Practices
- ✅ Be liberal at this stage — when in doubt, include
- ✅ Use dual screening with at least two independent reviewers
- ✅ Calculate inter-rater reliability (e.g., Cohen's kappa)
- ✅ Document reasons for exclusion
Full-Text Review
Obtain and read the full text of each paper that passed the title/abstract screen. Apply your inclusion and exclusion criteria more rigorously at this stage.
Key actions: Record specific reasons for each exclusion (you'll need these for your PRISMA flow diagram). Track how many papers were excluded at this stage and why.
Extract Data
Create a standardized data extraction form to collect key information from each included study. This typically includes:
Study Characteristics
- • Authors, year, country
- • Study design (RCT, cohort, etc.)
- • Sample size and demographics
Results Data
- • Primary outcomes and effect sizes
- • Secondary outcomes
- • Confidence intervals and p-values
Assess Quality & Risk of Bias
Evaluate the methodological quality and risk of bias in each included study using validated tools:
| Tool | Best For |
|---|---|
| Cochrane RoB 2 | Randomized controlled trials (RCTs) |
| Newcastle-Ottawa Scale | Observational studies (cohort, case-control) |
| ROBINS-I | Non-randomized interventional studies |
| AMSTAR 2 | Systematic reviews of systematic reviews |
Synthesize & Report
Combine your findings into a coherent narrative or statistical synthesis (meta-analysis). Follow the PRISMA 2020 guidelines for transparent reporting.
Narrative Synthesis
When studies are too heterogeneous for pooling. Organize by theme, population, or outcome.
Meta-Analysis
When studies are similar enough to combine statistically. Calculate pooled effect sizes with forest plots.
📊 Need a PRISMA diagram? Use our PRISMA 2020 flow diagram generator to create a publication-ready diagram in seconds.
Continue Learning
Ready to Start Your Systematic Review?
Lumina automates the most time-consuming part — screening. AI ranks papers by relevance so you find what matters faster.