The Review Protocol

Review Type

Systematic Literature Review: A Complete Guide

How to conduct a PRISMA 2020 systematic review: protocol, PROSPERO registration, exhaustive search, dual screening, risk of bias, and transparent reporting.

By Angel Reyes · Last updated

What is a systematic review?

A systematic review is a structured, reproducible synthesis of the evidence on a tightly defined question. It follows a pre-registered protocol, applies an exhaustive and documented search, screens studies against pre-specified inclusion criteria, assesses the risk of bias in each included study, and synthesizes the results — narratively, quantitatively, or both. Its defining features are transparency and reproducibility: another team, given the same protocol, should be able to arrive at the same set of included studies and broadly similar conclusions.

The modern systematic review traces its methodological lineage to Archie Cochrane's 1972 Effectiveness and Efficiency, which argued that medical practice should be grounded in a systematic appraisal of randomized trials. That argument led to the founding of the Cochrane Collaboration in 1993, and to the Cochrane Handbook for Systematic Reviews of Interventions, which remains the canonical methodological reference for reviews of healthcare interventions. Outside Cochrane, the Joanna Briggs Institute (JBI) publishes parallel methodologies for a wider range of questions, including qualitative, prevalence, and economic reviews.

The reporting standard for systematic reviews is PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses), published by Page and colleagues in BMJ and Systematic Reviews. PRISMA 2020 comprises a 27-item checklist and an updated flow diagram. Protocols are typically registered on PROSPERO, the international prospective register of systematic reviews maintained by the University of York's Centre for Reviews and Dissemination.

When to use a systematic review

A systematic review is the right choice when:

  1. You are answering a focused clinical, educational, or policy question. The question must be narrow enough that inclusion and exclusion criteria can be specified in advance. PICO, PICOS, or SPIDER frameworks help.
  2. A reproducible, defensible synthesis is required. Guideline developers, Cochrane reviews, HTA agencies, and many funders will accept nothing less.
  3. The evidence base is substantial but has not been synthesized recently. If the last high-quality systematic review is five or more years old and new studies have appeared, an update may be warranted.
  4. You have a team and a realistic timeline. A rigorous systematic review typically takes 12 to 18 months and involves at least two reviewers for screening and extraction.
  5. Quantitative pooling is possible or desirable. If studies report comparable effect sizes, you may extend the review into a meta-analysis.

Contrast with a scoping review (maps the landscape, does not assess bias), a narrative review (interpretive, no formal protocol), a rapid review (simplified methods for speed), and an umbrella review (synthesizes prior systematic reviews rather than primary studies).

Step-by-step process

Map your systematic review to the five review phases:

  1. Protocol and registration. Draft a protocol specifying your question, eligibility criteria, databases, search terms, screening procedures, extraction variables, risk-of-bias tool, and synthesis plan. Register on PROSPERO before you begin screening.
  2. Exhaustive search strategy. Search at least three major databases (e.g., MEDLINE via PubMed, Embase, CINAHL, or Scopus) using controlled vocabulary (MeSH, Emtree) and free-text synonyms joined with Boolean operators. Supplement with hand-searching key journals, citation chasing, and grey literature sources. Document every search string, date, and yield — a librarian co-author is invaluable here.
  3. Dual screening and selection. Deduplicate results, then screen titles/abstracts and full texts independently in duplicate (often using Covidence or Rayyan). Resolve disagreements by discussion or a third reviewer. Record reasons for full-text exclusions for the PRISMA flow diagram.
  4. Data extraction and risk of bias. Extract pre-specified variables (study design, population, intervention, comparator, outcomes, results) in duplicate using a piloted form. Assess risk of bias using Cochrane RoB 2 for randomized trials, ROBINS-I for non-randomized studies, or JBI critical appraisal tools as appropriate.
  5. Synthesis and reporting. Synthesize narratively (SWiM guidance), quantitatively (meta-analysis), or through structured approaches such as thematic synthesis. Grade the certainty of evidence using GRADE. Report in full compliance with PRISMA 2020, including the flow diagram and the 27-item checklist.

Reporting standards

PRISMA 2020 is the required reporting standard. Use the 27-item checklist (Page et al., BMJ 2021) and the updated flow diagram. Where applicable, add extension checklists: PRISMA-S for search reporting, PRISMA-P for protocols, PRISMA-NMA for network meta-analyses, and PRISMA-IPD for individual participant data reviews. Register the protocol on PROSPERO before screening begins; amendments after registration should be disclosed. See the full reporting standards guide for checklist mappings.

Common pitfalls

  • Vague eligibility criteria. If reviewers cannot apply a criterion consistently, the PRISMA flow diagram loses meaning. Pilot criteria on 50 records before the full screen.
  • Under-documented search strategy. PRISMA-S expects full search strings per database, date run, and record counts. Reconstructing this after the fact is painful and error-prone.
  • Single-reviewer screening or extraction. Without duplication, reviewers systematically miss studies and misread data. Budget time for dual review from the outset.
  • Treating risk of bias as optional. A systematic review without a formal bias assessment is closer to a narrative review. Use a validated tool and report results per study.
  • Over-generalizing from heterogeneous studies. Synthesis — especially pooled meta-analysis — requires comparable populations, interventions, and outcomes. When heterogeneity is high, prefer narrative synthesis with sub-group discussion.

Tools & templates

Use the PRISMA Flow Diagram Template to track identification, screening, eligibility, and inclusion counts. The Data Extraction Form Template pre-populates PICO fields, risk-of-bias ratings, and outcomes. Pair with the Inclusion/Exclusion Criteria Checklist and the Search Strategy Documentation Form. For deeper organization, the Literature Review Matrix consolidates extracted data across studies. All are available in the templates library.

Next steps

A systematic review is a project, not a paper. The studies that distinguish strong reviews from weak ones are the ones you find, extract, and reconcile in the first month — the protocol, the pilot screen, and the documented search. Build your extraction matrix once, rigorously, and every downstream step becomes easier. The Subthesis Literature Matrix gives your team a shared, structured workspace for extraction that maps cleanly onto PRISMA reporting requirements.