InsightMed Research Hub

How to Write a Systematic Review Protocol

InsightMed Research Hub · Published 19 November 2026 · 5-minute read

Short answer: A protocol is the plan for your review, written before you screen any studies. Follow the PRISMA-P 2015 checklist: state the question in PICO form, define eligibility criteria, list the databases and search strategy, describe how two reviewers will screen and extract data, name the risk of bias tool, and set out the analysis plan. Then register it on PROSPERO.
On this page
  1. Why a protocol matters
  2. The structure to follow
  3. Write the question first
  4. Eligibility criteria
  5. Search strategy
  6. Screening and data extraction
  7. Risk of bias and certainty
  8. Analysis plan
  9. Register it
  10. Common mistakes
  11. Frequently asked questions

Why a protocol matters

A protocol fixes your methods before you see the results. That protects the review from choices that are, consciously or not, shaped by the data. Journals and peer reviewers compare the published review with the protocol, and unexplained changes are one of the most common criticisms in review reports.

It also saves time. A team that agrees the eligibility criteria and extraction fields up front has far fewer disagreements during screening.

The structure to follow

Use the PRISMA-P 2015 checklist, which has 17 items. The main sections are:

  1. Administrative information: title that identifies the report as a protocol, registration number, authors and their contributions, amendments policy, and funding.
  2. Introduction: the rationale and the objective, written as an explicit question using PICO (population, intervention, comparator, outcome).
  3. Methods: eligibility criteria, information sources, search strategy, study records (data management, selection process, data collection), data items, outcomes and their priority, risk of bias, data synthesis, meta-bias and confidence in the cumulative evidence.

Write the question first

Everything else follows from the question. Write it in one sentence, then split it into PICO elements. For example: in adults after total thyroidectomy (P), does liquid or soft-gel levothyroxine (I) compared with tablets (C) change serum TSH levels (O)?

Decide your primary outcome now and name only one. Secondary outcomes can be several, but list them all. Outcomes added after you see the data look like selective reporting.

Eligibility criteria

Write criteria precise enough that two people would make the same decision on the same abstract. Cover study design, population, intervention and comparator, outcomes, minimum follow-up if relevant, language, and publication date limits. Give the reason for any restriction, for example why you exclude conference abstracts.

List every database (at minimum MEDLINE through PubMed, Embase and CENTRAL for intervention reviews), plus trial registries and reference list checking. Include the full search for at least one database in the protocol or its appendix, with both controlled vocabulary (MeSH, Emtree) and free text terms. A librarian review of the search, using the PRESS checklist, strengthens it.

Screening and data extraction

State that two reviewers will screen titles and abstracts and then full texts independently, how conflicts will be resolved, and which software you will use (for example Rayyan or Covidence). Do the same for data extraction, and list the data items you will collect. Plan a pilot of the extraction form on a few studies.

Risk of bias and certainty

Name the tool that fits each design: RoB 2 for randomised trials, ROBINS-I for non-randomised studies of interventions, QUADAS-2 for diagnostic accuracy studies, and the Newcastle-Ottawa Scale is widely used for observational studies. State that two reviewers will assess it independently. Say how you will rate the certainty of evidence, usually with GRADE.

Analysis plan

Describe the effect measure for each outcome (risk ratio, odds ratio, mean difference, standardised mean difference), the model (usually random effects) and the estimator, how you will assess heterogeneity (I² and tau²), and the software. List planned subgroup and sensitivity analyses with a reason for each, and say how you will assess small study effects (funnel plot and Egger test when there are at least 10 studies). Say what you will do if pooling is not appropriate.

Register it

Register the finished protocol on PROSPERO before data extraction starts, ideally before screening. Registration is free and gives a CRD number to cite in your paper. You can also publish the protocol in a journal such as Systematic Reviews or BMJ Open. See our guide on registering on PROSPERO.

Common mistakes

Sources

Moher D et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Systematic Reviews 2015;4:1. Shamseer L et al. PRISMA-P 2015: elaboration and explanation. BMJ 2015;349:g7647. Higgins JPT et al (eds). Cochrane Handbook for Systematic Reviews of Interventions.

Frequently asked questions

How long should a systematic review protocol be?

Most protocols are 2,000 to 4,000 words plus appendices. Length matters less than covering every PRISMA-P item clearly.

Can I change the protocol after registering it?

Yes. Update the registration record, date each change, and report every change and the reason for it in the final paper.

Do I need ethics approval for a systematic review protocol?

Usually not, because reviews use published data. Check your institution and target journal, as some ask for a formal exemption statement.

Do narrative reviews need a protocol?

A protocol is expected for systematic reviews and meta-analyses. Narrative reviews do not need one, though a short written plan still helps.