How to choose a meta-analysis service provider

Company, agency, firm, consultant or freelancer

The words people use for outside help with evidence synthesis overlap. A service provider is the general term for anyone who is paid to carry out part of the work. A company, agency or firm usually means an organization with several staff who share the work, so that a statistician, an information specialist and a writer may each handle their part. A consultant normally advises and may leave the work itself to the research team. A freelancer is an individual who takes on work directly. University statistical consultancies and library services also exist, and some offer limited help to members of the institution at no cost or low cost.

None of these forms is better in every case. A single experienced methodologist can be an excellent choice for a focused task, such as checking a statistical analysis. A team is more useful when a project needs several skills at once, or when it must continue if one person is unavailable. What matters is whether the provider can show how the work will be done and who will do it.

The rest of this guide applies to any of them. It is written from the point of view of the researcher who commissions the work and remains responsible for it.

Decide what you need before you ask

Most disappointing projects begin with an unclear brief. Before contacting a provider, write down the research question, the review type you think suits it, what already exists (a protocol, a list of studies, an extraction spreadsheet, a draft) and the output you need. Help is available at every stage, and the stage matters for the price and for the risk.

• The whole review: protocol, search, screening, extraction, risk of bias, analysis, certainty of evidence and manuscript. This is the largest commitment and the one in which clear contracts matter most.
• A single stage: for example the search strategy, data extraction or the statistical analysis. Check how the stage connects to the stages before and after it.
• The manuscript: editing, journal selection and submission, or responses to reviewers.

If you are unsure which of these you need, say so. A good provider will ask questions and propose a scope rather than quote immediately for something you may not need. The guide to review types may help you decide what kind of review fits your question.

Questions about method

The quality of an evidence synthesis is decided by its method, and a provider should be able to describe theirs in plain terms. Reasonable questions are the following.

1. Will there be a protocol, and will it be registered where that is appropriate? A protocol fixes the question, eligibility criteria, outcomes and analysis plan before the results are known. See PROSPERO registration.
2. How will the search be designed and documented? You should receive the full strategy for every database searched, with dates, so that the search can be repeated and reported. See search strategies.
3. Is screening and extraction done in duplicate, or checked by a second person? Errors in extraction are common and are a leading reason that pooled results cannot be reproduced. See data extraction.
4. Which tools assess risk of bias and certainty of evidence? The answer should name them, for example RoB 2 or ROBINS-I and GRADE, and explain how judgements are recorded.
5. How will the statistical methods be chosen? The choice of model, effect measure, heterogeneity estimator and handling of small studies should follow from the data and be justified in writing. See fixed-effect and random-effects models.
6. Will you receive the code and the data? Reproducible analysis means you can rerun it, and a reviewer or a co-author can check it. See reproducibility.
7. Which reporting guideline will be followed? For most systematic reviews the answer is PRISMA 2020 or one of its extensions.

Vague answers, or answers that avoid describing a method at all, are a reason to look elsewhere. A provider that cannot say how the work is done cannot show that it was done well.

Authorship, contribution and disclosure

The International Committee of Medical Journal Editors sets four criteria for authorship: a substantial contribution to the work, drafting or revising it critically, final approval of the version to be published, and agreement to be accountable for all aspects of it. A person who meets those criteria is an author. A person who helps without meeting them is acknowledged. Ghost-writing, in which a contributor is hidden from the journal, is not acceptable.

Ask the provider how contributions will be recorded. The answer should cover who will be named as an author or acknowledged, how the contributor roles will be described, and how any declaration the journal requires will be made. Ask also what the provider's policy is on the use of generative AI tools, since many journals require authors to disclose it, and compare the answer with the AI-use policy of any provider you consider.

The responsibility for the paper stays with the authors. A provider supports the work and does not take over the decisions. You should read, understand and be able to defend every statement in the manuscript that carries your name.

Promises that should concern you

Some claims are warning signs, because no honest provider can make them.

• A guarantee of acceptance or publication. Acceptance is decided by journal editors and reviewers. A provider can prepare a paper well, and can help you respond to reviewers, but cannot promise the result.
• A promise that the analysis will show a particular result. An analysis reports what the data show. A provider who offers a significant finding in advance is offering to choose the analysis to fit.
• Offers to arrange or buy authorship, to suggest reviewers who are friends of the author, or to influence an editor. These breach publication ethics, and journals and institutions treat them as misconduct. See the guidance of the Committee on Publication Ethics.
• Invented or untraceable references and data. Every study in a review should be one that exists, and every number should trace to its source.
• Unusually short timelines. A full systematic review takes months. A very fast turnaround usually means that steps such as duplicate screening or registration have been skipped. Where a shortened method is used on purpose, as in a rapid review, its limits should be stated.
• Pressure to pay everything in advance, or no written agreement.

Our own limits are set out in the research integrity statement and the terms, so that you can compare what any provider says in writing.

Confidentiality, data and contracts

An unpublished manuscript, a dataset or a protocol is the property of the research team. Before sending any of it, ask where it will be stored, who will have access, whether it will be entered into any external tool, and what happens to it when the project ends. A provider should be able to give clear answers and put them in the agreement.

The agreement itself should be written and should cover at least the following.

• The scope of the work and the deliverables, listed one by one.
• The timeline, with the points at which you need to supply or approve something.
• The fee, how it is calculated, when it is due, and what is charged if the scope changes.
• How many rounds of revision are included, and what happens after a journal rejection or a request for new analyses.
• Ownership of the data, the code and the manuscript, which should stay with the authors.
• Confidentiality, and the handling of any personal data under the rules that apply to you.
• How contributions will be acknowledged or declared.

Please do not send patient-identifiable information to any provider through a website form or by email. Describe the project in summary form first, and agree how files will be shared once the scope is settled.

A short comparison

The table is a general guide and not a ranking. Individual providers vary in every group.

How we approach it

My Meta-Analysis is a research-support service. We begin with a scoping conversation, agree the scope in writing, and work through a protocol, a documented search, screening, extraction, analysis, reporting and, where needed, journal submission and revision. The process page describes the stages, and the services are listed by stage and by type of paper. We do not guarantee acceptance by any journal, we do not alter data or results to fit a conclusion, and we do not provide clinical advice.

If you would like to discuss a project, send an enquiry with your question, the type of project and your target journal, and we will reply with the approach we would recommend and what we would need to begin.

Frequently asked questions