How long does a meta-analysis take, and what does it cost?

First-time authors tend to expect weeks and find that it takes months. Two studies of real projects give a fair picture. An analysis of 195 completed systematic reviews registered on PROSPERO found a mean of 67 weeks from registration to publication, with about five authors per review. An older analysis of 37 meta-analyses carried out by one company found a mean of about 1,100 hours of work, ranging from roughly 200 to 2,500, and the hours rose with the number of records retrieved.

Your project will land somewhere in those ranges, depending on the size of your search, the number of studies you include and how many people share the work. This guide shows what drives the time, a way to estimate your own, and what a meta-analysis costs apart from time.

What the figures say

Both studies are several years old, so treat them as a sense of scale rather than a forecast.

Calendar time: a mean of 67 weeks from registration to publication across 195 reviews (Borah and colleagues, BMJ Open, 2017). That includes the wait for a journal.
Team size: about five authors per review on average, with a wide spread.
How much is read and how little is kept: on average about 3% of the records retrieved were included, from searches that returned anything from 27 to over 92,000 records.
Hours of work: a mean of about 1,100 hours across 37 meta-analyses, ranging from roughly 200 to 2,500, and rising with the number of records retrieved (Allen and Olkin, JAMA, 1999).

What drives the time

Records to screen: every record is read, usually by two people.
Studies to include: each one needs extraction, a risk of bias judgement and checking.
Number of outcomes: each adds extraction and analysis.
Quality of reporting: missing or unclear data mean writing to authors and waiting for answers.
Full texts: papers you cannot open have to be requested or left out.
Team: more people share the reading but add coordination and the settling of disagreements.
Publication: peer review and revision can add months after the work is done.

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.

Estimate your own

Run your search first, or a draft of it, and count the records. Then multiply. As an illustration, suppose it returns 4,000 records and each title and abstract takes 30 to 60 seconds to read. One person screening all of them needs about 33 to 67 hours, and two people doing it independently need about 67 to 133 hours between them. Full-text review, extraction, risk of bias, analysis and writing come on top, each in proportion to the number of studies that pass screening.

Screening hours: records, times seconds per record, divided by 3,600, times the number of screeners.
Everything after screening: scale it with the number of studies you expect to include, and ask a similar review how long each step took.
A margin: add time for the unexpected, such as chasing missing data or a second round of searching.

What it costs apart from time

Amounts vary too much between institutions and topics to quote, but these are the items to budget for.

Information specialist: a librarian's time to build and check the search, which is worth paying for.
Database access: Embase and some other databases need a subscription, often through a university.
Full texts: papers behind a paywall, through interlibrary loan or purchase.
Translation: for studies in languages you do not read.
Statistics: a statistician's time, and software licences if you do not use free tools.
Publication: open-access fees, if the journal charges them.

Ways to shorten it without cutting corners

A narrow question with few outcomes.
A protocol that settles the methods before you start, so that no work is redone.
A well-built search that does not retrieve tens of thousands of irrelevant records.
Validated tools for the repetitive parts, such as ranking records for screening, with the checks described in AI screening for systematic reviews.

Sources

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

Borah R, Brown AW, Capers PL, Kaiser KA. Analysis of the time and workers needed to conduct systematic reviews of medical interventions using data from the PROSPERO registry. BMJ Open 2017;7:e012545.
Allen IE, Olkin I. Estimating time to conduct a meta-analysis from number of citations retrieved. JAMA 1999;282:634-635.

Questions

Can I do a meta-analysis in a month?

A rapid review with a very narrow question and a small search can sometimes be done in weeks. A full systematic review with a meta-analysis rarely can: the average in the PROSPERO analysis was 67 weeks, including publication.

How many hours does a meta-analysis take?

One older analysis of 37 meta-analyses reported a mean of about 1,100 hours, with a range of roughly 200 to 2,500. Your own estimate should start from the number of records your search returns.

Does AI make it faster?

It can shorten screening in particular, but you must test and check the tool, and that takes time too. The saving is largest when the search returns thousands of records.

How many people do I need?

The studies found an average of about five authors per review, and standards ask for at least two people to work independently on the key decisions. The guide on who you need for a meta-analysis sets out the roles.

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.