How to Write a Systematic Review Protocol (Step-by-Step)

A systematic review protocol is the plan you write before you start screening a single paper. It exists for one reason: to lock in your methodology before your findings can influence it. Reviewers and readers trust a systematic review more when they can see the plan was set in advance, not adjusted after the fact to fit what turned up.

Start with a well-formed question

Most protocols fall apart at the very first step because the research question is too vague to operationalize. "What is the evidence on diabetes treatment?" cannot be searched or screened consistently. Framing it with PICO — Population, Intervention, Comparison, Outcome — forces the specificity a search strategy actually needs. "In adults with type 2 diabetes (P), does metformin (I), compared to lifestyle intervention alone (C), improve HbA1c at 12 months (O)?" is searchable, screenable, and defensible.

Define eligibility criteria before you search

Inclusion and exclusion criteria belong in the protocol, not worked out during screening. Specify study designs you'll accept, population characteristics, publication date range, language restrictions if any, and outcome measures. Vague criteria are the single most common reason two reviewers disagree during screening — tightening this at the protocol stage prevents that disagreement from happening later, when it's expensive to resolve.

Build the search strategy

Your protocol should name every database you'll search, the exact search string or a representative version of it, and any supplementary methods — hand-searching reference lists, contacting authors for unpublished data, searching trial registries. A search strategy that can't be reproduced by someone else reading your protocol isn't finished yet.

Plan data extraction and synthesis in advance

Decide what data points you'll extract from each included study, and build the extraction form before you need it. Also decide, in the protocol, whether you expect to run a meta-analysis or a narrative synthesis, and under what conditions you'd choose one over the other (for example, sufficient clinical and statistical homogeneity across studies). Committing to this in advance protects you from the appearance of cherry-picking a synthesis method that flatters your results.

Register it

Once the protocol is written, register it — PROSPERO is the standard registry for health-related systematic reviews. Registration timestamps your methodology publicly, which is exactly the point: it's evidence you didn't adjust your approach after seeing the data. Many journals now require PROSPERO registration (or an equivalent) before they'll consider a systematic review for publication.

What reviewers actually check

When methods reviewers read a submitted systematic review, they compare the published methods section against the registered protocol looking for undisclosed deviations. Minor, justified changes are normal and expected — searches sometimes need adjustment once you see initial results. What draws objections is an unexplained mismatch between what was registered and what was actually done. Document any deviation and the reason for it as you go, rather than trying to reconstruct the justification later.

A protocol takes real time to do properly — usually more than researchers expect going in. But it's the single step most responsible for whether a systematic review holds up under peer review, and it's far cheaper to get right at the start than to defend after the fact.