SYSTEMATIC REVIEW: A PRACTICAL GUIDE (WITH META-ANALYSIS AND PRISMA).
What a systematic review actually is, how it differs from a narrative review (and why that difference isn't a matter of prestige), what the steps of the method are, when a meta-analysis is warranted, and what PRISMA is for. A guide written from the experience of working with authors, journals, and research teams.
IN THIS ARTICLE
By María García-Puente · Article · Updated July 2026
If you work in health research, sooner or later you'll run into the systematic review: because you want to publish one, because your academic work requires one, or because you need to read one critically before making a clinical decision. And that's where the problem we see most often in the journals and teams we work with shows up: many manuscripts are titled "systematic review" without having followed a systematic methodology. The title promises something the text doesn't deliver, peer review gets complicated, and the work loses credibility.
This guide explains what a systematic review actually is, how it differs from a narrative review (and why that difference isn't a matter of prestige), what the steps of the method are, when a meta-analysis is warranted, and what PRISMA is for. We wrote it from our experience supporting authors, journals, and research teams through this process.
What a systematic review is and what it's for
A systematic review is a structured, reproducible synthesis of the available evidence on a specific research question. The key word is reproducible: someone else, following the same steps you describe in your paper, should arrive at the same studies and equivalent conclusions. The whole method (a prior protocol, an exhaustive search, screening by two reviewers, a risk-of-bias assessment) exists for one reason: to minimize the bias of whoever conducts it.
What is it for? To answer a well-defined question as rigorously as possible when enough primary evidence already exists. That's why systematic reviews sit at the top of the evidence pyramid and underpin clinical practice guidelines, funding decisions, and health policy. It's also why they're among the most read and cited formats in biomedical journals: they save readers the work of finding, appraising, and synthesizing dozens of studies on their own.
The underlying methodological reference is the Cochrane Handbook for Systematic Reviews of Interventions, which sets out the international standard for the method.
Systematic review vs. narrative review: choosing wisely from the start
This is a good place to clear up a common misunderstanding: choosing between a narrative review and a systematic review isn't a matter of prestige — it's a matter of method and resources. A properly conducted systematic review is teamwork that takes months. A narrative review, by contrast, is a fully valid, more flexible format that a single author can carry out, one that shows the ability to synthesize and exercise independent judgment: an interpretive synthesis of the topic, carrying the voice and expert judgment of whoever writes it.
The key differences:
| Aspect | Narrative review | Systematic review |
|---|---|---|
| Purpose | Interpretive broad overview, with declared expert judgment | Structured, reproducible synthesis that answers a specific question while minimizing bias |
| Question | Broad and flexible; can evolve while writing | Defined in advance (PICO format), fixed in the protocol |
| Search | Selective; doesn't need to be exhaustive or reproducible | Exhaustive, reproducible, and dated, across multiple databases |
| Inclusion criteria | Implicit, at the author's discretion | Explicit and defined a priori |
| Bias assessment | Not mandatory | Mandatory for every included study |
| Team | Feasible for one person | At least three: two independent reviewers plus a third person to resolve disagreements |
| Reference guideline | SANRA as a quality guide | PRISMA + registered protocol (PROSPERO) |
The rule we apply as editors, and that we recommend to any author, is simple: if the title says "systematic review," it has to be one. If any of the method's requirements can't be guaranteed (independent double screening is the one that fails most often, because it requires a team), the more honest — and more solid — choice is to frame the work as a narrative review from the outset. Forcing the format usually ends up as a middle ground that's neither a good systematic review nor a good narrative one. And here's the good news: the literature search has to be done properly either way, so the effort you put into the search strategy is never wasted, whichever you decide.
The method, step by step
1. The question, in PICO format
It all starts with turning a clinical concern into an answerable question. The PICO format requires you to define the Population, the Intervention (or exposure), the Comparator, and the Outcomes. A well-defined question ("in adults with elevated lipoprotein(a), is the risk of acute myocardial infarction higher than in those with normal levels?") then shapes the eligibility criteria and the search itself. A vague question dooms the review before it starts.
2. Protocol and registration in PROSPERO
The protocol sets out, before you start, the question, the objectives, the eligibility criteria, and an initial approach to the search strategy. Registering it with PROSPERO, the international register of systematic review protocols in health, serves two purposes: it prevents duplicating reviews already underway, and it publicly records what you committed to doing, which protects against the temptation to adjust the criteria along the way when the results don't cooperate. More and more journals require it.
3. A reproducible search strategy
This is the information-retrieval core of the method, and where we add the most value as information specialists. A systematic search must be exhaustive, reproducible, and dated, run across multiple databases (usually PubMed/MEDLINE, Embase, and Cochrane, plus any subject-specific database if one exists), adapting the syntax to each one. Reproducible means, literally, that anyone should be able to copy and paste the strategy and get the same results.
In practice, this means combining thesaurus terms (MeSH in PubMed) with free text in the title and abstract, grouping the synonyms for each concept with OR, and combining the concepts with AND. It also means holding back: it's best not to add rigid qualifiers or filters (age, study type) at this stage, because they exclude relevant articles. It's better to retrieve a bit more and screen by reading title and abstract. The full strategy, with its dates, is published in the methods section or as supplementary material.
4. Screening by two reviewers
At least two reviewers independently read the title and abstract of every record retrieved, applying the eligibility criteria set out in the protocol. When they disagree, a third person decides. This step is the structural reason a systematic review can't be done alone: the independence of double screening is precisely what reduces selection bias, and there's no way to simulate it with a single person. Selected records then move on to full-text screening, following the same procedure.
5. Data extraction
From each included study, the relevant data is extracted in a standardized way, using a template defined in advance: population characteristics, intervention, comparator, outcomes, results. Working from a template stops each reviewer from recording different things and lays the groundwork for the synthesis.
6. Risk-of-bias assessment
Each included study is assessed using a quality-assessment tool suited to its design. This isn't about scoring for the sake of it: the judgment about risk of bias determines how much confidence the results deserve and how they should be interpreted in the synthesis. A review that adds up studies without weighing their quality is aggregating noise dressed up as evidence.
7. Synthesis
With all of the above in hand, the findings are synthesized. The synthesis can be qualitative (structured, but descriptive) or quantitative — and that's where meta-analysis comes in.
Meta-analysis: what it is and when it's appropriate
Meta-analysis and systematic review aren't synonyms, even though they're often used as if they were. Meta-analysis is a statistical technique that combines the numerical results of several studies to produce a pooled effect estimate, more precise than that of any single study. It's the quantitative synthesis that may — or may not — accompany a systematic review.
When is it appropriate? Only when the included studies are sufficiently homogeneous: comparable populations, interventions, and outcomes, measured in compatible ways. If the studies are too heterogeneous, combining them statistically produces a number that looks precise but has no real meaning; in that case, the right call is to stick with a qualitative synthesis and say so. A systematic review that doesn't include a meta-analysis isn't a flaw — often, it's a sign of sound methodological judgment. And when it does include one, the title should say so ("systematic review and meta-analysis"), because readers have a right to know what they'll find.
PRISMA and the flow diagram
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) is the international guideline that sets out how a systematic review should be reported. It's worth understanding its role clearly: PRISMA doesn't tell you how to conduct the review (that's what the Cochrane Handbook is for) — it tells you what information the manuscript needs to include so readers can appraise it and reproduce it. It consists of a 27-item checklist and the well-known flow diagram.
The PRISMA flow diagram captures, in a single figure, the full traceability of the process: how many records were identified in each database, how many duplicates were removed, how many were screened, how many were excluded (and why), and how many studies were ultimately included. It's the visual proof that the method was actually followed, and practically every biomedical journal requires it. If you can't fill in that diagram with real numbers once your review is finished, you have your diagnosis: what you did wasn't a systematic review.
Common mistakes
These are the ones we see most often, in journals and in academic work:
- Titling a piece "systematic review" when it isn't one. The most common and the most damaging: it creates expectations the text doesn't meet and undermines credibility with reviewers and readers.
- Starting to search without a protocol. Without criteria defined a priori, every inclusion decision gets made on the fly, and bias walks right in.
- A poor or non-reproducible search. Four terms in a single database, with no documented syntax or dates, won't hold up a systematic review.
- Filtering too early. Applying age, language, or study-type filters within the search itself excludes relevant articles that screening would have caught.
- Screening alone. Without two independent reviewers, there's no control over selection bias; it's a requirement of the method, not a nice-to-have.
- Skipping the risk-of-bias assessment, or treating it as a formality that doesn't shape the interpretation.
- Forcing a meta-analysis on heterogeneous studies, because a forest plot "looks better" than an honest qualitative synthesis.
- Thinking about visibility only at the end. A title and abstract written with the terms a reader would actually search for in PubMed mean the review, once published and through peer review, gets found and cited far more.
Learning the method before you need it
A well-conducted systematic review doesn't depend on a template or a tool — it depends on methodological judgment: knowing how to frame the question, build a reproducible search, organize the screening, and decide with good reason whether a meta-analysis is warranted. That judgment can be learned, and it's best learned before you're halfway through a manuscript.
We train researchers, librarians, and clinical teams in exactly this: advanced literature search, review methodology, and PRISMA-compliant writing, with real cases and the databases you'll actually use. If you have a review on the horizon, or a team that produces them regularly, take a look at our training courses and let's talk about what fits where you're starting from.
Reference sources: PRISMA 2020 Statement · Cochrane Handbook for Systematic Reviews of Interventions · PROSPERO, the international register of protocols
María García-Puente
Co-founder of AlterBiblio · Information specialist
KEEP READING
