Asked plainly, how long does a systematic review take? A full systematic review usually runs from several months to well over a year, with a common range of roughly six to eighteen months depending on the breadth of the question, the volume of records the search returns, the number of reviewers, and how much of the work is outsourced or shared. The timeline is driven less by the topic than by two labour-intensive stages, screening and data extraction, which scale with how many studies you find.
What actually drives the timeline
Most people underestimate a review because they picture the writing, which is the smallest part. The clock is really set by the size of the evidence and the number of hands available. A narrow question that returns a few hundred records and has two committed reviewers can move quickly; a broad question that returns tens of thousands of records, or a single reviewer trying to do everything alone, can stretch the project past a year. Understanding which stage consumes the time is the key to planning a realistic schedule.
A realistic timeline, stage by stage
Protocol and question (two to six weeks)
Framing the question with a focused review question and writing the protocol is fast in calendar terms but pivotal: a sloppy question inflates every later stage. Allow extra days if you register on PROSPERO and wait for it to be processed.
Searching (two to four weeks)
Building, testing, and running a reproducible search strategy across multiple databases takes a few weeks, longer if you add grey literature searching or have the strategy peer-reviewed first. The search itself is quick to run once designed; the design is where the care goes.
Screening (one to four months)
This is usually the single longest stage. Two reviewers must independently screen every title and abstract, then every full text, against the eligibility criteria. The duration scales almost directly with record count, which is why deduplication and a tight question pay off so heavily here. Using screening tools can shave weeks but does not remove the dual-reviewer requirement.
Data extraction and risk of bias (one to three months)
Pulling consistent fields from every included study with a piloted extraction process, ideally in duplicate, runs alongside formal risk-of-bias assessment. The more studies survive screening, the longer this takes, which is another reason a precise question saves months downstream.
Synthesis and writing (one to three months)
If you pool results in a meta-analysis, the analysis and forest plots add time but are not the bottleneck. Writing up against PRISMA 2020, including the flow diagram and full search, then revising for submission, rounds out the project.
A worked example: how record count sets the clock
Numbers make the relationship concrete. Suppose your search returns 4,000 unique records after de-duplication. A practised reviewer screens roughly 120 to 150 titles and abstracts per hour at this stage, so one pass takes about 30 hours. Because screening is done in duplicate, that is 60 reviewer-hours before any conflict resolution. If each reviewer can spare ten focused hours a week, title and abstract screening alone runs to roughly three calendar weeks, and that is before the slower full-text stage. Now double the records to 8,000 and the same arithmetic doubles the calendar. This is why a tightly framed question and disciplined removal of duplicate records save more time than any other single decision.
Full-text screening is slower per item but smaller in volume. If 350 records survive to full text and each takes ten to fifteen minutes to read and judge against the eligibility criteria, that is another 60 to 90 reviewer-hours in duplicate. The lesson is that you cannot estimate a timeline until the search has run, because the record count is the input every later stage multiplies.
Where reviews quietly lose months
The published estimates assume work proceeds without interruption, which it rarely does. The hidden delays are predictable and worth budgeting for:
- Registration waits. The administrative check when you register a protocol on PROSPERO takes time before screening should begin, and that wait is dead calendar if you have not planned around it.
- Inter-library loans. Full texts that are not retrieved immediately stall the eligibility phase, and chasing a single hard-to-find report can add a fortnight.
- Missing data requests. Contacting study authors for an unreported standard deviation often means waiting weeks for a reply, or none at all.
- Conflict resolution. A low level of agreement between screeners, measured with Cohen’s kappa, signals ambiguous criteria and forces a re-screen that doubles the work for the affected records.
- Peer review and revision. Once submitted, journal turnaround and a round of revisions routinely add several months that sit entirely outside the team’s control.
How to make a review faster without cutting corners
Three levers move the timeline honestly. First, narrow the question so the search returns fewer records, because everything downstream scales with that number. Second, add reviewers to the screening and extraction stages, since these parallelise well. Third, prepare and pilot your forms and search before you start, so the slow stages run cleanly the first time. If speed is the priority and some rigour can be traded, a streamlined rapid review format may be the better fit. For the full sequence behind these estimates, see the stages of a systematic review.
What does not work is the most common shortcut: a single reviewer rushing screening alone. Beyond reintroducing selection bias, it removes the parallelism that genuinely shortens the calendar, because one person cannot screen two records at once. A meta-analysis, by contrast, is rarely the place to economise: the pooling itself is fast once the data are clean, and our statistics service can run it in days rather than weeks. The honest planning rule is to estimate from the record count once the search has run, add a contingency for the hidden delays above, and resource the two slow stages first.