Systematic review statistics in one paragraph: roughly 80 systematic reviews are published every day, a twenty-fold increase since 2000 (Hoffmann et al., 2021). The average review takes 67.3 weeks to complete and publish (Borah et al., 2017), involves a median of 5 authors (Page et al., 2016), and costs an estimated $141,195 in researcher time when conducted in-house (Michelson and Reuter, 2019). Meanwhile, 46.4 percent of registered reviews in one large cohort were never published (Runjic et al., 2019). Every figure on this page comes from a named, linked, peer-reviewed source, so you can verify and cite each one directly.
- Nearly 80 systematic reviews were published per day in 2019, up from about 4 per day in 2000. Indexed systematic reviews rose from 1,432 in 2000 to 29,073 in 2019, a twenty-fold increase. Source: Hoffmann F, Allers K, Rombey T, et al. Journal of Clinical Epidemiology, 2021. PubMed 34091022
- 58.9 percent of systematic reviews include a meta-analysis (Hoffmann et al., 2021, same source).
- Systematic review output grew 2,728 percent between 1991 and 2014, against 153 percent growth for PubMed overall. Ioannidis called the volume "epidemic proportions." Source: Ioannidis JPA. Milbank Quarterly, 2016. PubMed 27620683
- Mean time to complete and publish: 67.3 weeks (about 15.5 months), from an analysis of 195 completed reviews registered in PROSPERO. Source: Borah R, Brown AW, Capers PL, Kaiser KA. BMJ Open, 2017. Full text at PubMed Central
- Median time from registration to publication: 1.6 years (interquartile range 0.9 to 2.1) in a 2024 cohort. Source: Burgos MA et al. Cochrane Evidence Synthesis and Methods, 2024. PubMed 40474875
- Individual participant data meta-analyses typically take 2 to 3 years from protocol to publication, per Cochrane's own methods group. Source: Cochrane IPD Meta-analysis Methods Group
- Cochrane requires searches to be rerun within 12 months before publication (MECIR standard C37, mandatory), which is why slow reviews pay a re-screening penalty. Source: Cochrane MECIR Manual
- The fastest documented full systematic review took 61 person-hours over 12 calendar days, using the accelerated "2-week systematic review" methodology. Source: Clark J, Glasziou P, Del Mar C, et al. Journal of Clinical Epidemiology, 2020. PubMed 32004673
- Across ten accelerated reviews, the median was 11 workdays (range 5 to 20). Source: Scott AM, Glasziou P, Clark J. Journal of Clinical Epidemiology, 2023. PubMed 36898508
- Cochrane's rapid review guidance targets completion within 6 months. Source: Garritty C et al. Journal of Clinical Epidemiology, 2021. Full text
- Funded reviews take longer and use bigger teams than unfunded ones: mean 42 versus 26 weeks of active work, and 6.8 versus 4.8 team members (Borah et al., 2017, same source as above).
For a phase-by-phase breakdown of where those months go, see our guide to how long a systematic review really takes.
- Mean 5 authors per review (standard deviation 3), from the PROSPERO cohort (Borah et al., 2017).
- Median 5 authors, interquartile range 4 to 6, from a random sample of 300 systematic reviews. Source: Page MJ et al. PLOS Medicine, 2016. Full text
- Cochrane will not publish a single-author review; teams must include systematic review methodology and statistical expertise. Source: Cochrane Handbook, Chapter 1
- Accelerated review teams ran at a median of 6 people (Scott et al., 2023).
- Search yields range from 27 to 92,020 records per review, and on average only 2.94 percent of screened records make it into the final review (Borah et al., 2017).
- An experienced reviewer screens about 2 abstracts per minute. Source: Wallace BC, Trikalinos TA, Lau J, et al. BMC Bioinformatics, 2010. Full text
- A costed workload model puts title and abstract screening at 1.0 minute per record, with full double screening of a large review costing up to 75,139 pounds and text mining cutting the title and abstract workload by more than 60 percent. Source: Shemilt I, Khan N, Park S, Thomas J. Systematic Reviews, 2016. Full text
- Recent reviews screened between 458 and 5,471 references each and included 5 to 81 studies (Scott et al., 2023).
- The median review includes 15 studies (Page et al., 2016).
Dual screening is the single largest workload block in these numbers, which is why we built AI-assisted literature screening with a validated 99.1 percent recall workflow, and why our free PRISMA 2020 flow diagram generator exists for the reporting side.
- An in-house systematic review costs an estimated $141,194.80 in researcher time: 1.72 person-years at the United States Bureau of Labor Statistics average medical scientist salary of $82,090. Source: Michelson M, Reuter K. Contemporary Clinical Trials Communications, 2019. Full text
- The same study estimates large pharmaceutical companies spend $16.8 million per year and large academic institutions $18.7 million per year on systematic literature reviews (roughly 119 and 132 reviews per year respectively).
- Screening alone costs between 37,860 and 75,139 pounds for a large review, depending on whether single screening with text mining or full double screening is used (Shemilt et al., 2016).
Our full breakdown of what commissioning a review actually costs, including where professional teams save researcher hours, complements these institutional figures. If you are weighing the in-house route against commissioned systematic review support, the 1.72 person-year figure is the honest baseline to compare against.
Publication and quality benchmarks
- 46.4 percent of pain systematic reviews registered in PROSPERO were never published, and only 16.3 percent of registry records had an accurate status. Source: Runjic E, Rombey T, Pieper D, Puljak L. Journal of Clinical Epidemiology, 2019. PubMed 31476376
- A 2024 South American cohort found only 36.9 percent of registered reviews reached publication in indexed journals (Burgos et al., 2024).
- Only 16 percent of published reviews had a publicly accessible protocol, and just 29 percent reported using a reporting guideline such as PRISMA (Page et al., 2016).
- PRISMA 2020 adherence runs at about 72.8 percent of items in journals that require the checklist, versus 68.4 percent where they do not. Source: Suda C et al. Cureus, 2025. PubMed 40027066
- When retracted trials are removed from meta-analyses, statistical significance changes in 11 percent of them, and effect estimates shift by at least 10 percent in 42 percent of primary-outcome analyses. Source: Grana Possamai C, Cabanac G, Perrodeau E, et al. JAMA Internal Medicine, 2025. Full article
- Over 94 percent of citations to retracted biomedical papers never mention the retraction, and among systematic reviews citing retracted trials, 22 percent had pooled the retracted data. Source: Bakker C, Boughton S, Faggion CM, et al. BMJ Evidence-Based Medicine, 2023. Full text
Quality failures of this kind are method failures, not writing failures, which is where meta-analysis support from our statisticians most often gets called in: retraction checks, dual extraction, and risk-of-bias discipline are what the benchmarks above say most reviews skip.
These benchmarks are compiled and verified from the primary sources linked above; each figure links to the study that produced it. If you reference this compilation, cite the underlying study for the specific number, or link this page for the collection: Research Gold, "Systematic Review Statistics: Verified Benchmarks," 2026.
Figures were verified against the full text of each source in July 2026. Where studies disagree (for example, mean completion time of 67.3 weeks versus a median of 1.6 years from registration to publication), both are reported with their cohorts, since they measure different populations and different start points. Figures from sources that could not be independently fetched and verified were excluded.