How Long Does a PRISMA Systematic Review Take? (By Stage)

Dr. Rajni Garg
Executive Lead, Research & Content Strategy
6 min read
A systematic review reported to PRISMA standards usually takes 12 to 18 months from registration to publication. The best-known estimate, from a study of reviews registered in PROSPERO, is an average of 67.3 weeks, with individual reviews taking anywhere from 6 to 186 weeks. The average team had five authors.
Speed is possible, though. One team finished a full systematic review in 12 calendar days, using 61 person-hours and a set of automation tools. Most reviews land between those extremes. Where yours falls depends on the size of your search, the size of your team and whether you run a meta-analysis, all of which you can plan for.
TL;DR
Plan for 12 to 18 months from protocol to publication. The published average is 67.3 weeks.
PRISMA adds little time. It is a reporting guideline, a 27-item checklist and a flow diagram, not a separate stage of work.
Screening and data extraction take about 60% of the calendar. Project management takes more time than most teams expect.
Search size, the number of reviewers and a meta-analysis drive the timeline more than anything else.
Automation and AI tools can compress the work, as a 2-week review showed, but only with a tight question and an experienced team.
Is PRISMA the method or the paperwork?
The paperwork, mostly. PRISMA 2020 tells you what to report: a 27-item checklist and a flow diagram that shows how many records you found, screened, excluded and included. The time goes into the review method underneath, usually set out in a handbook such as Cochrane’s or JBI’s.
That matters for planning. Filling in the checklist and drawing the PRISMA flow diagram takes hours if you record your numbers as you go. Rebuilding those numbers at the end turns it into days.
What the published timelines show
Study | What it measured | Result |
|---|---|---|
Borah et al., 2017, BMJ Open | Reviews registered in PROSPERO, from registration to publication | Mean 67.3 weeks, range 6 to 186 weeks, mean of 5 authors |
Clark et al., 2020, Journal of Clinical Epidemiology | One review done with automation tools and a dedicated team | 61 person-hours over 12 calendar days, 71 person-hours to the accepted version |
Nussbaumer-Streit et al., 2021, Journal of Clinical Epidemiology | Scoping review of 34 studies on resource use | Study selection, data extraction and critical appraisal were the most resource-intensive steps |
Borah’s data also show how uneven the workload is. The smallest search in the sample returned 27 records and the largest 92,020. On average, only 2.94% of the records a search retrieved made it into the final review. A review with 500 records to screen and one with 15,000 are different projects, even under the same PRISMA heading. Funded reviews also had bigger teams, a mean of 6.8 people against 4.8 for unfunded ones.
Where the time goes, stage by stage
No study times every stage of every review, but the American Institutes for Research’s MOSAIC guide gives a practical split for planning a meta-analysis. Apply its shares to Borah’s 67-week average and you get a rough calendar.
Stage | Share of total time (AIR MOSAIC) | Rough weeks at a 67-week pace |
|---|---|---|
Protocol and project management | 5% | About 3 |
Literature search | 5% | About 3 |
Title and abstract screening | 15% | About 10 |
Full-text screening | 20% | About 13 |
Data extraction (coding) | 25% | About 17 |
Data cleaning and meta-analysis | 15% | About 10 |
Writing up and dissemination | 15% | About 10 |
Treat the last column as an illustration, not a forecast. It still teaches something useful: screening and extraction take around 60% of the calendar. Time saved there matters far more than time saved on the protocol. The AIR guide’s own advice for the analysis stage is blunt: “Plan for more time than you think!”
Nussbaumer-Streit’s review adds a point most timelines leave out. Project management and administration took a large share of review time in the studies it found. Meetings, resolving screening conflicts and chasing full texts all belong in your plan.
What makes a review take longer?
Search size. The more records you retrieve, the longer screening takes. Every extra thousand records costs reviewer hours.
Dual screening and extraction. Two independent reviewers are standard, and they need time to resolve disagreements.
Full-text retrieval. Papers behind paywalls, in other languages or only in print slow things down.
A meta-analysis. Pooling results adds data cleaning, statistics and sensitivity analyses.
Team availability. Most reviewers fit the work around other jobs, so a project measured in person-hours can still stretch over months.
Peer review and revisions. Borah’s figure runs to publication, so journal turnaround is part of the 67 weeks.
How do teams finish faster?
The 2-week review proves the timeline is not fixed. According to the GRADE Working Group’s write-up, the team screened 1,381 deduplicated records down to 8 included studies. They held short daily meetings, blocked protected time for each member, and resolved screening conflicts straight after screening. They also leaned on automation: a word frequency analyzer to pick search terms, SRA Helper to screen faster and find full texts, and RobotReviewer to help with risk of bias.
Few teams will match that pace, but the same levers work at any speed.
Narrow the question before you search. A precise PICO question cuts the number of records you screen. The PICO tab in the AnswerThis search string builder helps you set out each element.
Build and test the search string early. Check it finds a few studies you already know belong in the review.
Pilot your screening criteria on 50 to 100 records, so both reviewers apply them the same way.
Screen in software, not spreadsheets. Our roundup of systematic review software compares the main options.
Record your PRISMA numbers at every stage, so the flow diagram is a five-minute job at the end.
Consider AI with human oversight. The 2025 joint position statement from Cochrane, the Campbell Collaboration, JBI and the Collaboration for Environmental Evidence says that in rapid reviews AI might act as a second screener, as long as its use is declared and checked.
If time is the real constraint, ask whether you need a full systematic review at all. A scoping review maps a field rather than answering a narrow question about effectiveness, and it may suit your aim better.
A realistic plan for a student review
For a master’s or PhD student working with a part-time second reviewer, this is realistic for a review with a few thousand records:
Month 1: protocol and registration.
Month 2: build, test and run the searches.
Months 3 to 5: title, abstract and full-text screening.
Months 6 and 7: data extraction and risk of bias assessment.
Months 8 to 10: synthesis and writing.
That is eight to ten months before submission, which fits inside Borah’s average once journal review is added. For the method behind each stage, see our guide on how to conduct a systematic literature review, or start with what a systematic review is if you are still deciding.
Where AnswerThis fits
AnswerThis tools take the slow, repeatable work out of three stages.
Search. Describe your topic or PICO question in the AnswerThis search string builder and it drafts a Boolean search you can edit, formatted for the database you pick from 12 options such as PubMed, Scopus, ERIC and Google Scholar. Synonyms come grouped in OR blocks with concepts joined by AND, so testing it against your known studies takes minutes. Three runs per browser are free.
Risk of bias. The risk of bias assessment tool drafts an assessment of each study with RoB 2, ROBINS-I, QUADAS-2, the Newcastle-Ottawa Scale or a JBI checklist, quoting the text behind each answer for your reviewers to confirm.
Reporting. Enter the counts you logged into our PRISMA flow diagram generator. It lays them out as a PRISMA 2020 diagram, with a version adapted for scoping reviews, and you can download it as a PNG, an SVG, an editable PowerPoint slide or JSON. It is free and needs no account.

To accelerate your workflow, explore Systematic Review Tutorial: How to Do a Full Review With AI, Step by Step (PRISMA, Meta-Analysis).
FAQs
Can you do a systematic review in 3 months?
It is possible with a narrow question, a small evidence base and at least two committed reviewers. Clark et al. (2020) finished in 12 calendar days, but they were experienced reviewers with protected time and daily meetings. For most people, 3 months is tight and only works if the search returns a manageable number of records.
How long does a systematic review take for a PhD student?
Often 8 to 12 months as one chapter of a thesis, and sometimes longer if it includes a meta-analysis. Build in time for a second reviewer, since most methods guides expect independent dual screening.
Does following PRISMA make a review take longer?
Very little. PRISMA tells you what to report. The extra time comes from the systematic methods themselves, such as a comprehensive search and dual screening.
Which stage of a systematic review takes the longest?
Screening and data extraction. In the AIR planning guide they take about 60% of the total time, and Nussbaumer-Streit’s scoping review found study selection, data extraction and critical appraisal to be the most resource-intensive steps.
Is a scoping review faster than a systematic review?
Often not by much. Scoping reviews skip formal risk of bias assessment, but their broad questions can return large numbers of records to screen.
Plan around the biggest stages, and start with a search you have tested. Draft yours in the search string builder, then log your counts for the PRISMA flow diagram generator as you go.



.png)
