The protocol is the step most people want to rush, and it is the one that decides whether your review holds up. It is the plan you commit to before you look at a single result: the question, the search, the rules for including studies, and how you will analyze them. Write it well and the rest of the review runs on rails. Skip it, or write it loosely, and every later decision becomes a place where your own preferences can quietly steer the outcome.
That is the whole point of a protocol. It locks in your choices while you are still neutral, before you know which studies would give you the answer you were hoping for.
A systematic review protocol is a document that specifies, in advance, exactly how the review will be conducted. It states the question, the eligibility criteria, the databases and search terms, the screening and extraction process, the risk-of-bias tool, and the planned synthesis. It is written before searching begins and is meant to be followed as written.
Think of it as a pre-registration of your intentions. If you later deviate from it, and reviews sometimes do for legitimate reasons, you disclose the change and why. The protocol turns "trust me, I was objective" into "here is the plan I committed to before I saw the data."
The value is bias prevention. Once you have seen your search results, it is human nature to nudge the criteria toward a cleaner story. Maybe you narrow the population to exclude the studies that muddy the picture, or you add an outcome that happens to show a strong effect. None of this feels dishonest in the moment. A committed protocol removes the temptation, because the rules were set before you had anything to gain from bending them.
There are practical reasons too. Journals increasingly require a registered protocol. Registration prevents duplicate reviews and stakes your claim to a question. And writing the protocol forces you to think through the entire project before you invest months in it, which is often where you discover the question is too broad or the evidence too thin.
Before you search. This is not negotiable if you want the review to count as genuinely systematic. Piloting a quick scoping search to gauge whether enough evidence exists is fine and often sensible. Running your full, documented search before the protocol is finalized is not, because at that point your criteria can be shaped by what you have already seen.
A complete protocol covers the following. The PRISMA-P guideline (PRISMA for Protocols) sets out these items in full, and following it helps you avoid gaps.
Title and registration. A clear title identifying it as a systematic review protocol, plus registration details once you have them.
Background and rationale. Why the review is needed, what is already known, and what gap it addresses. Enough context to justify the question without turning into a full literature review.
Objectives and question. The specific question, usually framed with PICO (Population, Intervention, Comparison, Outcome) or a variant suited to your review. This is the anchor for everything else.
Eligibility criteria. Exactly which studies are in and which are out: population, intervention, comparator, outcomes, study designs, publication dates, languages, and any other filters. Precise criteria are what make screening reproducible.
Information sources. Which databases you will search, along with other sources such as trial registries, grey literature, and reference-list checking.
Search strategy. Your actual search terms and how they combine, ideally the full strategy for at least one database. This is what makes the search reproducible, so do not leave it vague.
Study selection process. How screening will work: how many reviewers, whether they screen independently, and how disagreements are resolved.
Data extraction. What data you will collect from each study and how, including whether extraction is done in duplicate.
Risk-of-bias assessment. Which appraisal tool you will use, matched to your study designs (for example RoB 2 for randomized trials, ROBINS-I for non-randomized ones).
Synthesis and analysis plan. How you will combine the findings: narratively, or with a meta-analysis, and if the latter, which model and how you will handle heterogeneity and subgroups. Planning subgroup analyses in advance matters, because post-hoc subgroups are easy to fish for significance in.
Certainty of evidence. Whether and how you will rate certainty, commonly with GRADE.
Registration makes your protocol public and time-stamped. For health-related reviews, the main registry is PROSPERO, which is free and widely recognized. You submit your protocol details and receive a registration number that you cite in the final manuscript.
Registration does two things. It proves your plan predated your results, which is your defense against accusations of shaping the review after the fact. And it prevents wasted effort, since others can see that your review is underway. Some fields and journals accept other registration routes or the publication of the protocol as a standalone article, but for most health and clinical reviews, PROSPERO is the expected home.
Protocols are commitments, not prisons. Sometimes you discover a genuine reason to deviate: an outcome that turns out to be unmeasurable, a subgroup that no study reported. When that happens, the rule is simple. Make the change, document it clearly in the final report, and explain why. Undisclosed deviations are the problem, not deviations themselves. A transparent, justified change preserves the integrity of the review; a silent one destroys it.
Writing it after searching. This defeats the entire purpose. The protocol only prevents bias if it predates the results.
Vague eligibility criteria. "Relevant studies on the topic" is not a criterion. If two reviewers cannot apply your rules and reach the same decision, the rules are too loose.
Skipping the search strategy. A protocol that names databases but omits the actual terms is not reproducible. Include the real strategy.
No analysis plan. Deciding how to synthesize after you see the data invites cherry-picking. Commit to your approach, including any subgroup analyses, in advance.
Never registering. An unregistered protocol offers far weaker protection against the suspicion that you adjusted the plan to fit the findings.
Do I have to register my protocol? It is strongly recommended and required by many journals. Registration on PROSPERO or an equivalent time-stamps your plan and is your main defense against accusations of altering the review after seeing results.
What is the difference between a protocol and the review? The protocol is the plan, written before searching. The review is the completed work that follows that plan. The protocol says what you intend to do; the review reports what you did and found.
Can I change my protocol after registering? Yes, when there is a legitimate reason. You must disclose the change and explain it in the final report. Undisclosed deviations undermine the review's credibility.
What is PRISMA-P? PRISMA-P is the reporting guideline for systematic review protocols. Following its checklist helps ensure your protocol includes every element a reader and reviewer would expect.
How long should a protocol be? Long enough to be complete and reproducible, which usually means several pages. The search strategy and eligibility criteria alone take real space if done properly.
A systematic review protocol is the plan you commit to before you can be tempted to bend it: the question, the criteria, the search, and the analysis, all fixed while you are still neutral. Write it before you search, cover the PRISMA-P items, register it on PROSPERO or an equivalent, and disclose any later changes honestly.
It is the least glamorous step and the one that does the most to keep your review trustworthy. Get it right and everything downstream has a firm foundation. For the full sequence that follows, see our guide on how to conduct systematic review?.
Ready to move from protocol to execution? Verflux runs the search, screening, appraisal, and meta-analysis your protocol lays out, in one place.
All analytical features included in the free trial. No credit card, no installation, no R or Python.
Create free account