RevMan is the systematic review software developed by Cochrane for preparing, maintaining, and updating Cochrane systematic reviews, with built-in support for meta-analysis forest plots, risk-of-bias tables, and GRADE summary of findings. The acronym stands for Review Manager, the family name has been the Cochrane Collaboration standard since the 1980s, and the platform now exists in two generations: the legacy RevMan 5 desktop application and the current cloud-based RevMan Web. Understanding which generation to use, what RevMan does natively, and where its limits sit is essential for any review team working inside the Cochrane workflow and useful even for teams considering RevMan as a starting platform for a non-Cochrane review.
RevMan Web and RevMan 5: Two Generations of One Workflow
RevMan 5 is the desktop application that Cochrane authors used for nearly two decades. It runs as a standalone program on Windows, macOS, and Linux, stores reviews as proprietary .rm5 files, and is locally installed on the reviewer's machine. RevMan 5 reached end of active development in 2020, although authors with legacy reviews can still install it and Cochrane continues to provide compatibility tooling.
RevMan Web is the cloud-based successor that Cochrane now expects all new Cochrane reviews to use. RevMan Web runs in the browser, stores the review on Cochrane servers, supports multi-author collaboration with real-time editing, integrates with Cochrane's central systems for editorial management and publication, and includes updated analytical features such as native support for the RoB 2 risk-of-bias tool. The migration from RevMan 5 to RevMan Web has been the largest single platform change in the history of Cochrane review production.
For new reviews the choice is settled: use RevMan Web. For legacy reviews, the choice depends on whether the existing .rm5 file has been migrated through Cochrane's conversion tool and whether the editorial group has signaled a hard cutover. Most active Cochrane Review Groups have completed the transition.
Origins and the Cochrane Standardization Argument
The original Review Manager dates to the late 1980s and grew out of the Nordic Cochrane Centre's effort to standardize the data structures, statistical procedures, and visual outputs of the systematic reviews that the Cochrane Collaboration was beginning to publish at scale. Before RevMan, each Cochrane review team prepared meta-analyses in whichever software was available locally, which produced reviews that varied in figure style, data table format, and statistical defaults.
RevMan's standardization mandate is the single most important feature to understand. Every Cochrane review opened in RevMan, regardless of topic area or author team, has the same structural skeleton: an outline of background, methods, results, discussion, and references; a defined set of comparisons each containing one or more outcomes; data extraction tables in a fixed grid; standard forest plot conventions; a uniform Summary of Findings table; and journal-ready output formatted to The Cochrane Database of Systematic Reviews style.
This standardization is invisible to authors who have only used RevMan, who experience it simply as the way reviews are written. It becomes very visible to authors who try to migrate non-Cochrane reviews into RevMan or vice versa. The standardization is also what makes RevMan unsuitable for some advanced analytical workflows that require custom plots, non-standard effect measures, or methodological extensions that the platform does not anticipate.
Core Features You Get Out of the Box
RevMan, in either generation, ships with a set of features that cover the methodological backbone of a standard healthcare systematic review:
Data extraction tables with fixed columns for study identification, methods, participants, interventions, outcomes, and notes. The tables are linked to the analysis backend so that any cell change propagates to the forest plot.
Automatic forest plots for fixed-effect and random-effects meta-analysis using the Mantel-Haenszel, inverse variance, or Peto methods, with subgroup analysis, heterogeneity statistics, and the standard Cochrane visual conventions.
Risk-of-bias tables for the Cochrane Handbook reference standard tools. RevMan 5 supported the original Cochrane risk-of-bias tool. RevMan Web supports RoB 2 for randomized trials, ROBINS-I for non-randomized studies, and QUADAS-2 for diagnostic accuracy studies as part of the platform.
Summary of Findings tables that integrate with the GRADE approach, presenting absolute and relative effects with certainty ratings and footnotes in the format expected by Cochrane reviewers.
Characteristics of included studies tables that summarize study design, participants, interventions, outcomes, and notes in the structured format Cochrane reviews require.
PRISMA flow diagram integration that aligns the study selection narrative with the PRISMA 2020 reporting requirements, though many teams build the diagram in a graphics application and import the final image.
Reference management through built-in handling of citation metadata, deduplication, and inclusion-exclusion tracking, though most teams use a separate reference manager such as EndNote or Zotero for primary deduplication.
What RevMan Does Well
RevMan's strongest feature is consistency. Every output looks like every other Cochrane review, which is exactly what Cochrane authors and editors want. A reviewer who has worked through one Cochrane review can navigate any other Cochrane review's RevMan file without orientation, which is a methodological advantage no general-purpose statistics package can match.
RevMan also does standard meta-analysis methods well. Mantel-Haenszel, inverse variance, and Peto methods are implemented correctly with appropriate continuity corrections and heterogeneity statistics. Forest plots respect Cochrane visual conventions, including weight columns, summary diamonds at the bottom of subgroups and overall, and clearly labeled axis ticks.
The team-collaboration features in RevMan Web are a step change over RevMan 5. Multiple authors can edit a review simultaneously, the platform tracks individual contributions for authorship transparency, and the cloud-based storage removes the version-merging headaches that plagued the desktop generation. For multi-site review teams, RevMan Web is qualitatively easier to manage than RevMan 5.
RevMan's editorial integration is also worth noting. Submitted Cochrane reviews flow directly from RevMan Web into the Cochrane editorial system, which reduces the formatting overhead that journal submission typically involves. For Cochrane authors, this integration alone justifies the platform.
What RevMan Does Not Do
RevMan's narrow methodological scope is the most important limitation to plan around. The platform does not natively support several methods that are standard in non-Cochrane meta-analysis:
Network meta-analysis is not supported in RevMan. Cochrane authors who run network meta-analyses use external tools such as R's netmeta package, GeMTC, or WinBUGS, then import the resulting figures and tables as static images into RevMan.
REML estimation of the between-study variance is not implemented. RevMan uses the DerSimonian-Laird estimator by default, which performs poorly when the number of studies is small. Reviewers who want REML or other modern variance estimators must run the analysis externally.
Prediction intervals are not generated automatically. For a comprehensive prediction-interval treatment, reviewers run the analysis in R using metafor or in Stata using meta and report the prediction interval as a supplementary figure.
Bayesian meta-analysis with custom priors is not supported. Bayesian methods are increasingly important in evidence synthesis, particularly for rare-event meta-analyses and for network meta-analysis, and require external tooling.
Custom plots beyond the standard forest plot template are not supported. Galbraith plots, Baujat plots, GOSH plots, leave-one-out sensitivity plots, and contour-enhanced funnel plots all require external graphics software.
Qualitative synthesis in the systematic review sense, including meta-ethnography, framework synthesis, and thematic synthesis, is not supported by RevMan. RevMan is fundamentally a quantitative-synthesis platform.
Individual participant data meta-analysis with one-stage models, two-stage models with custom random effects, or treatment-effect interactions is not supported. IPD reviews use Stata or R for the core analysis and use RevMan only for the aggregate sensitivity analyses if at all.
This list looks long, but for most healthcare intervention reviews running standard aggregate meta-analyses with two-arm comparisons, RevMan's feature set is sufficient. The limitations matter most for methodologically ambitious reviews and for non-Cochrane reviews that want to use more flexible methods.
