Generate professional forest plots for your meta-analysis in minutes. No coding, no installation, no cost. Enter data, customize, and export.
A forest plot (sometimes called a blobbogram) is the standard graphical display used in meta-analysis to present the results of multiple independent studies side by side. First introduced by Lewis and Clarke in 1991, the forest plot has become an indispensable tool in evidence-based medicine, appearing in virtually every published systematic review and meta-analysis.
The forest plot earns its name from the visual impression it creates: a "forest" of horizontal lines (confidence intervals) with squares (point estimates) scattered across the display. Each line represents one study, and taken together, they provide an immediate visual summary of whether studies agree, how precise each estimate is, and what the overall combined effect looks like.
Every forest plot contains the same core components. Understanding these elements is essential for both creating and interpreting forest plots correctly.
Let us break down each element:
| Element | Visual | Meaning |
|---|---|---|
| Point Estimate (Square) | Black square on each row | The effect size calculated from that individual study (e.g., odds ratio, risk ratio, mean difference). The center of the square marks the exact value. |
| Confidence Interval (Horizontal Line) | Line extending from the square | The 95% confidence interval around the point estimate. A wider line means less precision (more uncertainty). If the line crosses the null line, the result is not statistically significant for that individual study. |
| Weight (Square Size) | Larger or smaller squares | Reflects the relative contribution of each study to the pooled estimate. Studies with larger sample sizes and smaller standard errors receive more weight. The percentage is typically shown in the rightmost column. |
| Pooled Estimate (Diamond) | Diamond at the bottom | The combined effect across all studies. The center of the diamond is the pooled point estimate. The width of the diamond represents its 95% confidence interval. This is the main result of the meta-analysis. |
| Line of No Effect | Vertical line at OR/RR = 1.0 or MD = 0 | The reference line where there is no difference between intervention and control. Studies or the pooled estimate crossing this line are not statistically significant. |
| Heterogeneity Statistics | Text below the plot | I², Q statistic, tau², and associated p-values that quantify how much the studies disagree with each other beyond what is expected by chance. |
A forest plot is far more than a decorative figure in a paper. It is the primary communication vehicle for meta-analytic results, and there are several reasons why it remains irreplaceable nearly four decades after its introduction.
A single forest plot conveys information that would otherwise require paragraphs of text and multiple tables. In one image, a reader can see the effect size, precision, weight, and statistical significance of every included study plus the overall result. For busy clinicians, policymakers, and journal reviewers who may not read every line of your results section, the forest plot is often the first and most thoroughly examined element of a meta-analysis.
Forest plots place every study on the same scale with the same axis, making direct comparison straightforward. A reader can instantly identify which studies found the largest effects, which found the smallest, and which found effects in the opposite direction. This visual alignment eliminates the cognitive load of comparing numbers scattered across a table.
Before even looking at the I² statistic, you can gauge heterogeneity visually from a forest plot. If the confidence intervals of all studies overlap substantially and the squares cluster tightly, heterogeneity is likely low. If the lines are spread across the plot with minimal overlap, heterogeneity is likely high. This visual impression is often more intuitive than a numerical I² value, especially for readers who are not biostatisticians.
A forest plot makes it immediately obvious when one study stands apart from the rest. An outlier will have its square and confidence interval far from the cluster of other studies. This visual identification is critical because a single outlying study can substantially shift the pooled estimate, and readers and reviewers will want to know whether such influential studies were investigated through sensitivity analysis.
The varying sizes of the squares provide an immediate sense of which studies contributed the most to the pooled result. A very large square indicates a study with substantial weight (typically a large trial with narrow confidence intervals), while a tiny square indicates a study with little influence on the pooled estimate. This weight information helps readers assess whether the overall result is driven by one dominant study or represents a balanced combination.
When studies are grouped by a categorical variable (e.g., study region, dose level, patient age group, risk of bias rating), forest plots can display separate pooled diamonds for each subgroup. This subgroup forest plot allows readers to visually compare effects across subgroups and identify potential effect modification -- a critical part of exploring heterogeneity.
Reading a forest plot can seem intimidating the first time, but the process becomes straightforward once you understand the components. Follow these steps systematically.
Look at the x-axis label to determine what effect measure is being used. Common options include:
The vertical line represents the value of "no effect." For OR and RR, this is 1.0. For MD and SMD, this is 0. Any study or pooled estimate whose confidence interval crosses this line has not found a statistically significant result at the conventional p < 0.05 level.
Go through each row from top to bottom. For each study, note:
Forest plots typically label the left and right sides of the plot (e.g., "Favours Treatment" on the left and "Favours Control" on the right, or vice versa). Always confirm which direction favors the intervention being studied. This labeling convention varies between reviews and software tools, so never assume.
The diamond at the bottom is the headline result of the meta-analysis. Examine three things:
Below the forest plot, look for heterogeneity statistics:
Step back and look at the overall pattern of the forest plot:
Researchers have several options for generating forest plots. The best choice depends on your technical skill level, budget, and specific needs. Below is a detailed comparison of the most widely used tools.
| Feature | MetaReview | R / metafor | RevMan 5 | Stata (metan) | Excel |
|---|---|---|---|---|---|
| Cost | Free | Free (open source) | Free (Cochrane authors) | $295-$895/year | Microsoft 365 subscription |
| Coding Required | No | Yes (R programming) | No | Yes (Stata syntax) | No (but manual) |
| Installation | None (browser-based) | R + packages | Desktop software | Desktop software | Desktop/web |
| Effect Sizes | OR, RR, MD, SMD | All (highly flexible) | OR, RR, RD, MD, SMD | All (flexible) | Manual calculation |
| Forest Plot Quality | Publication-ready | Highly customizable | Publication-ready | Publication-ready | Not publication-ready |
| Subgroup Analysis | Yes | Yes | Yes | Yes | No |
| Funnel Plot | Yes | Yes | Yes | Yes | Manual only |
| Sensitivity Analysis | Yes (leave-one-out) | Yes (all types) | Limited | Yes | No |
| Export Format | SVG, PNG | PDF, SVG, PNG, TIFF | Image, PDF | PDF, PNG, EPS | Image only |
| Learning Curve | 5 minutes | Days to weeks | 1-2 hours | Days to weeks | Hours (tedious) |
| Best For | Beginners, fast results | Power users, custom analysis | Cochrane reviews | Academic statisticians | Not recommended |
MetaReview is a free, browser-based meta-analysis tool designed for researchers who need publication-quality results without the steep learning curve of R or Stata. You enter your data, choose your effect size and model, and the tool generates a forest plot automatically. The entire process takes under 10 minutes. MetaReview is ideal for medical researchers, graduate students, and anyone conducting their first meta-analysis.
The R statistical language, combined with the metafor package by Wolfgang Viechtbauer, is the gold standard for advanced meta-analysis. It supports every effect size measure, every model specification, meta-regression, multivariate meta-analysis, and nearly infinite plot customization. However, it requires R programming knowledge, which represents a significant barrier for researchers without a statistics or programming background. If you are comfortable with code, metafor is extraordinarily powerful.
RevMan 5 (and the newer RevMan Web) is developed by the Cochrane Collaboration and is the default tool for Cochrane systematic reviews. It has a clean interface, supports common effect sizes, and generates forest plots that meet Cochrane formatting standards. However, it requires a free Cochrane account, desktop installation, and is less flexible than R for non-standard analyses. RevMan Web is gradually replacing RevMan 5 with a browser-based experience, but it is still less feature-rich than MetaReview for non-Cochrane users.
Stata with the community-contributed metan (or meta in Stata 16+) commands is widely used in epidemiology and health economics. It produces high-quality forest plots and supports advanced methods. The main drawback is cost: a Stata license ranges from $295 to $895 per year depending on the edition. Like R, it requires familiarity with its command-line syntax.
While it is technically possible to create something resembling a forest plot in Microsoft Excel, it requires manually calculating effect sizes, constructing error bars, formatting the chart, and the result is almost never publication-ready. Journal reviewers and editors will immediately recognize an Excel-generated forest plot, and it will not meet the presentation standards expected in peer-reviewed publications. We strongly recommend against using Excel for forest plots when purpose-built tools exist for free.
This section walks you through the complete process of generating a publication-ready forest plot using MetaReview's free online tool. The entire workflow takes approximately 5 to 10 minutes.
Before opening the tool, organize your extracted data. You need one row per study with the following information:
| Column | Description | Example |
|---|---|---|
| Study Name | First author and year | Smith 2019 |
| Events (Intervention) | Number of events in treatment group | 23 |
| Total (Intervention) | Total participants in treatment group | 150 |
| Events (Control) | Number of events in control group | 45 |
| Total (Control) | Total participants in control group | 148 |
| Column | Description | Example |
|---|---|---|
| Study Name | First author and year | Jones 2020 |
| Mean (Intervention) | Mean value in treatment group | -2.4 |
| SD (Intervention) | Standard deviation in treatment group | 1.8 |
| N (Intervention) | Sample size of treatment group | 85 |
| Mean (Control) | Mean value in control group | -0.6 |
| SD (Control) | Standard deviation in control group | 2.1 |
| N (Control) | Sample size of control group | 82 |
The choice of effect size determines how your forest plot presents results. This decision should be made during your protocol development, not after data extraction.
Open MetaReview in your web browser. No account registration is needed. Navigate to the data entry section and choose your data type (binary or continuous).
For each included study, enter the study name and its corresponding data values. MetaReview validates your inputs in real time -- it will flag impossible values (e.g., events exceeding the total sample, negative standard deviations, or missing required fields) so you can correct errors immediately.
Alternatively, click the CSV import button to upload a prepared spreadsheet. MetaReview accepts standard CSV format with headers matching the expected column names. This is the fastest method when you have more than 5-6 studies.
Navigate to the results page. MetaReview automatically computes the pooled effect size, 95% confidence interval, and heterogeneity statistics based on your selected model.
Switch to the forest plot view. MetaReview renders a clean, publication-ready forest plot showing each study's effect estimate, confidence interval, weight, and the pooled diamond. The heterogeneity statistics (I², Q, p-value) are displayed below the plot.
Review the plot carefully. Make sure the axis labels are correct, the direction labels (e.g., "Favours Treatment" and "Favours Control") match your comparison, and the study order makes sense. Customize the appearance if needed (see the Customization section below).
Before exporting, interpret what the forest plot tells you:
MetaReview also generates an auto-written results paragraph that you can use as a starting point for your manuscript's results section. This paragraph includes the pooled estimate, confidence interval, p-value, heterogeneity statistics, and model used.
Click the download button to export your forest plot. MetaReview supports SVG format (vector graphics, resolution-independent, ideal for journal submission) and PNG format (raster graphics, suitable for presentations and web use). For journal submission, always choose SVG unless the journal specifically requires a different format.
If your journal requires a specific format (EPS, TIFF, PDF), export as SVG first and then convert using free tools like Inkscape. SVG preserves full quality at any conversion target resolution.
While MetaReview generates a clean, publication-ready forest plot by default, several customization options are available to match your preferences or specific journal requirements.
Control the order in which studies appear in the forest plot:
If you have defined a subgroup variable (e.g., study region, intervention dose, risk of bias level), MetaReview can generate a subgroup forest plot. This displays studies grouped by the subgroup variable, with a separate pooled diamond for each subgroup and an overall pooled diamond at the bottom. Subgroup forest plots are essential for exploring heterogeneity and are frequently requested by journal reviewers.
MetaReview provides options to adjust the visual appearance of your forest plot:
For customization beyond what MetaReview offers natively, export your forest plot as SVG and open it in a vector graphics editor:
This two-step workflow (generate in MetaReview, fine-tune in a vector editor) gives you both the statistical accuracy of a dedicated meta-analysis tool and the visual polish of a professional design application.
Creating a forest plot is only half the task. Correctly interpreting and reporting what the forest plot shows is equally important. This section covers the key elements you need to address in your manuscript.
The pooled diamond is the primary result. Report the pooled effect size, its 95% confidence interval, and the p-value. For example:
Remember that statistical significance does not equal clinical significance. A pooled OR of 0.98 (95% CI: 0.97-0.99) is statistically significant but may have negligible clinical impact. Always discuss the magnitude of the effect in relation to clinical relevance thresholds established in your field.
Report heterogeneity statistics alongside the pooled estimate. The key metrics are:
| Metric | What It Measures | Interpretation |
|---|---|---|
| I² | Percentage of variation due to true heterogeneity | 0-40% low, 30-60% moderate, 50-90% substantial, 75-100% considerable (Cochrane guidelines). Note the overlapping ranges reflect that interpretation depends on context. |
| Cochran's Q | Whether observed differences exceed sampling error | A significant Q (p < 0.10) indicates heterogeneity. Q has low statistical power with few studies, so use a lenient alpha of 0.10 rather than 0.05. |
| Tau² (tau-squared) | Estimated between-study variance | Expressed in the same squared units as the effect size. Tau (the square root) is more interpretable as the standard deviation of the distribution of true effects. |
| Prediction Interval | Range of effects expected in future studies | Wider than the confidence interval. If the prediction interval crosses the null, the effect may not be present in all settings, even if the pooled estimate is significant. |
While the confidence interval tells you the uncertainty around the average pooled effect, the prediction interval tells you the range within which the true effect of a future study is expected to fall. In a random-effects meta-analysis, the prediction interval is always wider than the confidence interval because it incorporates both the uncertainty of the mean estimate and the between-study variability (tau²).
For example, a pooled RR of 0.70 (95% CI: 0.55-0.90) looks impressive, but if the 95% prediction interval is 0.30-1.65, it means the effect in a future study could plausibly favor either the intervention or the control. This additional information helps clinicians and policymakers understand how generalizable the finding is.
If your forest plot includes subgroups, compare the pooled diamonds across subgroups. Key questions to address:
Even experienced researchers make errors when creating and presenting forest plots. Avoiding these common mistakes will improve the credibility and accuracy of your meta-analysis.
Selecting an inappropriate effect size measure is one of the most consequential errors. Using an odds ratio when a risk ratio is more clinically meaningful, or using a mean difference when studies measured outcomes on different scales (requiring SMD), will produce misleading results. The effect size choice should be justified in your protocol and should match both your data type and the clinical question.
Combining studies that measure fundamentally different outcomes into a single forest plot produces a meaningless pooled estimate. For example, pooling a study that measures "complete remission" with one that measures "partial response" using the same binary framework treats these as equivalent endpoints when they are not. Similarly, mixing adjusted and unadjusted effect sizes without accounting for the difference introduces bias. Ensure outcome definitions are sufficiently comparable before pooling.
Reporting a clean pooled estimate while ignoring an I² of 85% is a red flag for any reviewer. When heterogeneity is substantial (I² > 50%), you must explore its sources through subgroup analysis, meta-regression, or sensitivity analysis. Simply reporting the random-effects model pooled estimate and moving on suggests you either do not understand the heterogeneity or are choosing to ignore it. If the heterogeneity cannot be explained, acknowledge it transparently and discuss its implications for the certainty of your findings.
A beautiful, significant forest plot can be completely misleading if it is based on a biased set of studies. Small studies with null or negative results are less likely to be published, leading to an overestimation of the true effect. Always generate a funnel plot alongside your forest plot and perform quantitative tests (Egger's test, Begg's test) when you have 10 or more studies. If publication bias is suspected, apply the trim-and-fill method or conduct a sensitivity analysis excluding studies with high risk of bias.
Using a fixed-effect model when studies clearly come from different populations with different true effects will underestimate the uncertainty of the pooled estimate (confidence intervals will be too narrow). Conversely, using a random-effects model with very few studies (k < 5) can produce unreliable tau² estimates. The model choice should be justified based on clinical and methodological judgment, not based on which model produces a more favorable result.
A single transposed digit can dramatically alter a study's effect size and, by extension, the pooled estimate. Common errors include swapping intervention and control columns, entering events instead of non-events, mixing up mean and standard deviation values, and using the wrong decimal point. Always double-check your data against the original source publications. MetaReview flags impossible values (events exceeding totals, negative SDs), but plausible but incorrect values will not be caught automatically.
A forest plot in a manuscript should always be accompanied by adequate text explaining what is shown, what effect measure is used, what model was used, and what the heterogeneity statistics indicate. Dropping a forest plot into a paper without interpretation forces readers to draw their own conclusions, which may be incorrect. Every forest plot deserves a dedicated paragraph in your results section.
Running both fixed-effect and random-effects models and reporting whichever produces the more favorable result is a form of outcome reporting bias. The analysis plan, including model choice, should be prespecified in your protocol. If you report both models for transparency (which is acceptable), clearly designate one as the primary analysis and the other as a sensitivity check.
The final step is exporting your forest plot in a format that meets journal submission requirements. Here is what you need to know about file formats, resolution, and common journal expectations.
| Feature | SVG (Vector) | PNG (Raster) |
|---|---|---|
| Resolution | Infinite (scales to any size) | Fixed (depends on export size) |
| File Size | Small (text-based XML) | Larger (pixel-based) |
| Editable | Yes (in Illustrator, Inkscape) | Limited (pixel editing only) |
| Best For | Journal submission, archival | Presentations, web, email |
| Print Quality | Always perfect | Depends on DPI (need 300+) |
| Conversion | Easily converts to PDF, EPS, TIFF | Cannot upscale without quality loss |
While specific requirements vary by journal, most biomedical journals follow similar guidelines:
A well-written figure caption is part of the publication process. Include these elements:
MetaReview is a free online forest plot generator. Go from raw data to a publication-quality forest plot in under 10 minutes. No installation, no coding, no cost.
Open MetaReview - It's FreeGet notified about new features, forest plot tips, and meta-analysis guides.
No spam. Unsubscribe anytime.
A forest plot is the standard graphical display used in meta-analysis to visualize the results of multiple studies investigating the same question. Each study is represented as a horizontal row containing a square (the point estimate or effect size) and a line extending from it (the 95% confidence interval). The size of each square is proportional to the study's statistical weight in the pooled analysis -- larger squares represent studies that contribute more to the overall result. At the bottom of the plot, a diamond shape represents the pooled (combined) effect estimate across all studies. A vertical line marks the "null effect" (1.0 for ratio measures like odds ratio and risk ratio, or 0 for difference measures like mean difference). Forest plots allow researchers, clinicians, and reviewers to see at a glance whether the included studies agree with each other, how precise each estimate is, and whether the overall combined effect is statistically significant.
Reading a forest plot follows a systematic process. First, identify the effect measure on the x-axis (OR, RR, MD, or SMD) and locate the vertical line of no effect. Each row represents one study: the square is the point estimate and the horizontal line is the 95% confidence interval. Larger squares indicate studies with more weight. If a study's line crosses the null line, that study alone did not find a statistically significant result. Next, look at the diamond at the bottom, which shows the pooled estimate. The diamond's center is the pooled effect size and its width is the pooled confidence interval. If the diamond does not touch or cross the null line, the overall result is statistically significant. Finally, check the heterogeneity statistics below the plot (I-squared and Q p-value) to assess how consistent the studies are. An I-squared above 50% suggests substantial heterogeneity, meaning the studies may not be measuring exactly the same thing, and further investigation through subgroup analysis is warranted.
The diamond at the bottom of a forest plot represents the pooled (combined) effect estimate from the meta-analysis. It is the summary result that synthesizes all included studies. The center of the diamond indicates the pooled point estimate -- for example, an overall odds ratio of 0.65 or an overall mean difference of -2.3. The left and right tips of the diamond represent the lower and upper bounds of the 95% confidence interval around the pooled estimate. A narrow diamond means the pooled estimate is precise (there is little uncertainty about the overall effect), while a wide diamond means there is substantial uncertainty even after combining multiple studies. The critical question is whether the diamond overlaps the vertical line of no effect. If the diamond falls entirely to one side of the null line without touching it, the pooled result is statistically significant at the p < 0.05 level. In subgroup analyses, you may see multiple diamonds, one for each subgroup, with a final overall diamond at the very bottom.
Forest plots and funnel plots are both standard figures in meta-analysis, but they serve entirely different purposes. A forest plot displays the results: it shows each study's effect size, confidence interval, and weight, plus the pooled estimate. It answers the question "what is the overall effect and how do individual studies compare?" A funnel plot, in contrast, is a diagnostic tool for detecting publication bias. It plots each study's effect size on the x-axis against a measure of precision (usually standard error) on the y-axis. In the absence of publication bias, the points should form a symmetric, inverted funnel shape centered on the pooled estimate. Asymmetry in the funnel plot may indicate that small studies with non-significant or unfavorable results were not published, leading to a biased set of included studies. However, funnel plot asymmetry can also result from genuine heterogeneity, methodological differences between small and large studies, or statistical artifact. Formal tests like Egger's regression test or Begg's rank correlation test complement the visual inspection of funnel plots.
MetaReview is a free, browser-based forest plot generator that requires absolutely no coding, programming, or software installation. You simply open the MetaReview website in any modern web browser (Chrome, Firefox, Safari, Edge), enter your study data through a point-and-click interface, select your effect size type (OR, RR, MD, or SMD) and analysis model (fixed-effect or random-effects), and the tool automatically generates a publication-ready forest plot. You can customize the appearance, add subgroups, and export the result as SVG or PNG. The entire process takes 5 to 10 minutes. Unlike R (which requires installing R, RStudio, and packages like metafor or meta, then writing code), or Stata (which requires a paid license and knowledge of command syntax), MetaReview eliminates all technical barriers. It is designed specifically for medical researchers, graduate students, and anyone conducting a meta-analysis who prefers a visual, no-code workflow.
Yes. MetaReview offers several customization options to tailor your forest plot's appearance. You can adjust the color scheme to match your preferences or differentiate subgroups, choose how studies are sorted (by effect size, weight, year, or entry order), toggle the display of study weight percentages and numerical confidence interval values, and modify axis labels to accurately describe your comparison (e.g., "Favours Drug A" vs. "Favours Placebo"). The default styling follows established academic conventions for clean, professional output that journals expect. For additional customization beyond what the tool offers natively -- such as adding annotations, changing specific fonts, or adjusting element positions -- you can export the forest plot as an SVG file and edit it in free vector graphics software like Inkscape, or in professional tools like Adobe Illustrator or Affinity Designer. This gives you full control over every visual element while maintaining the statistical accuracy of the original plot.
MetaReview exports forest plots in SVG (Scalable Vector Graphics) format, which is the ideal format for journal submission because SVG is resolution-independent -- it scales to any size without any quality loss. Most biomedical journals accept SVG directly, or accept PDF and EPS, which can be easily generated from SVG using free tools like Inkscape (File > Save As > PDF or EPS). For journals that specifically require TIFF format at 300-600 DPI, open the SVG in Inkscape, set the document size to match the journal's column width (typically 85mm for single column or 170mm for double column), and export as TIFF at 600 DPI (File > Export PNG Image, set DPI to 600, then convert PNG to TIFF). MetaReview also offers direct PNG export for presentations and web use. The key principle is: always start with SVG and convert down to raster formats as needed, never the other way around. This ensures maximum quality at every step.
The best free forest plot tool depends on your technical background and specific needs. For researchers who want publication-ready results without any coding, MetaReview is the top choice: it runs entirely in the browser, supports all major effect sizes (OR, RR, MD, SMD), provides both fixed-effect and random-effects models, generates forest plots with automatic weight calculation and heterogeneity statistics, offers subgroup analysis, sensitivity analysis (leave-one-out), funnel plots, and exports in SVG format. For researchers comfortable with programming, R with the metafor package (by Wolfgang Viechtbauer) offers the most powerful and flexible meta-analysis capabilities, including meta-regression, multivariate meta-analysis, and virtually unlimited plot customization. RevMan (Review Manager) from the Cochrane Collaboration is free for Cochrane-affiliated authors and is the standard for Cochrane systematic reviews, though it requires desktop installation and is less flexible for non-Cochrane analyses. For the vast majority of researchers conducting standard meta-analyses, MetaReview provides the best balance of capability, ease of use, and zero cost.