Meta-analysis protocol checklist: what to decide before you search

A protocol is the plan you commit to before you see the results. It stops the review drifting towards whatever looks interesting in the data, it lets others see that the topic is taken, and it makes the later write-up much faster, because the methods section is already written.

The PRISMA-P 2015 checklist lists 17 items for a protocol. This guide groups what to decide into a shorter working checklist, adds the analysis choices that people most often leave open, and points to registration.

Why write it first

Outcomes, comparisons and models chosen after seeing the data can be bent, without anyone meaning to, towards a result. A dated protocol shows that your choices came first. PRISMA 2020 asks reviews to give their registration details or say there were none, to say where the protocol can be found, and to describe and explain any changes (items 24a to 24c).

The checklist

Title and registration: say that it is a protocol, and give the registration number once you have one.
Team: who does what, and who is the contact person.
Rationale and question: why the review is needed, and the question as population, intervention, comparator and outcome.
Eligibility: study designs, population, interventions and comparators, outcomes, time period, languages and publication status.
Sources and search: the databases and registers, other sources such as reference lists, and a draft search strategy for at least one database.
Selection: how many reviewers screen, whether independently, how disagreements are settled, and any software or AI tool.
Data collection: the form, what is extracted, who extracts it, and how missing data are handled.
Outcomes: the primary outcome, the secondary outcomes and the time points, ranked in advance.
Risk of bias: the tool for each study design, and how the judgements will be used in the analysis.
Synthesis plan: the effect measure, the model and estimator, how heterogeneity is assessed, and what you will do if the studies cannot be pooled.
Pre-specified analyses: subgroup and sensitivity analyses, with the reason for each, kept few.
Reporting bias: how you will look for missing results, remembering that funnel plot tests are only for 10 or more studies.
Certainty of evidence: usually GRADE, and the outcomes it will cover.
AI and automation: any tool you plan to use, for what, and how you will check it.
Amendments: how changes will be recorded, with the date and the reason.
Funding and interests: who pays, and any conflicts.

Let SynthGap run the first pass

Enter an area. SynthGap proposes candidate topics, checks each against existing systematic reviews, and lists the primary studies behind it. Start on the Free plan: no card needed, and every run shows every topic it finds.

The choices people leave open

Fixed-effect or random-effects: decide before you see the data, and give the reason. Do not choose after seeing which gives the result you wanted.
Which outcome is primary: more than one primary outcome multiplies the chance of a chance finding.
The subgroups: a long list of subgroup analyses will turn up something by chance.
What counts as too different to pool: write the rule down before you extract.
How you will handle multi-arm trials and several reports of one trial.

Register it and keep it findable

Register the protocol where reviewers can find it: PROSPERO for reviews with a health-related outcome, or the Open Science Framework for other reviews. Journals such as Systematic Reviews also publish protocols, which gives you a citable record.

Sources

The figures and rules on this page come from these sources.

Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ 2015;350:g7647.
Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71.
Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ 2011;343:d4002.
Cochrane Handbook for Systematic Reviews of Interventions, chapter 13: assessing risk of bias due to missing evidence in a meta-analysis.

Questions

Is a protocol required for a meta-analysis?

Not by law, but good practice and most reporting guidance expect one, and many journals ask whether the review was registered. Cochrane reviews require a published protocol.

Can I change my protocol after registering?

Yes, if you record the change, the date and the reason, and report it in the review. What you cannot do is change it silently after seeing the results.

Where do I publish a protocol?

You can register it, on PROSPERO or the Open Science Framework, publish it in a journal that accepts protocols, or do both. The registration number goes in the abstract and the methods of the final review.

How long should a protocol be?

Long enough that someone else could repeat your methods. The 17 items of PRISMA-P are a good minimum, and the draft search strategy and the synthesis plan are the parts worth the most care.

Keep reading

General methodological guidance, not medical advice. SynthGap's suggestions are AI-generated from the published literature, so verify them before registering a protocol.