Create publication-ready forest plots from your meta-analysis data with configurable confidence levels (80%, 90%, 95%, 99%), prediction intervals, and your choice of DerSimonian-Laird or REML tau-squared estimator. Explore cumulative meta-analysis, Galbraith (radial) plots, and Baujat diagnostic plots alongside leave-one-out sensitivity analysis and automatic subgroup detection with between-group Q-test. Import via CSV/Excel or the inter-tool pipeline, customize font sizes, copy plots to clipboard, auto-generate a methods paragraph, and export reproducible R metafor code. All inputs auto-save so you never lose work.
Move data between tools automatically. Compute effect sizes, then send results to Forest Plot, Funnel Plot, or Heterogeneity analysis with one click.
No data in pipeline yet. Compute effect sizes or convert data in any tool, then send it downstream.
Drag & drop a file or
CSV, TSV, Excel (.xlsx/.xls) - max 500 rows
Load sample data to see how the tool works, or clear all fields to start fresh.
| Study | Effect | CI Lower | CI Upper | Subgroup | |
|---|---|---|---|---|---|
Import studies directly from the Effect Size Calculator or other upstream tools via the inter-tool pipeline, drag-drop a CSV or Excel file with fuzzy column matching, or enter each study manually with its name, effect size, and confidence interval bounds.
Select fixed-effect or random-effects model with your choice of DerSimonian-Laird or REML tau-squared estimator. Pick a confidence level (80%, 90%, 95%, or 99%) and enable prediction intervals for random-effects models.
Set the effect measure label, line of no effect value, sort order, and font size (8-16px). Assign studies to subgroups for sub-pooled estimates and automatic between-group Q-test.
Switch between the forest plot, cumulative meta-analysis, Galbraith (radial) plot, and Baujat plot tabs. Run leave-one-out sensitivity analysis to identify influential studies.
Download the plot as PNG or SVG, copy it to clipboard with one click, export data as CSV or Excel, auto-generate a publication-ready methods paragraph, or copy reproducible R metafor code.
Forward your forest plot results to the Funnel Plot tool for publication bias assessment or to the Heterogeneity Calculator for deeper analysis, all with one click from the inter-tool pipeline.
Need this done professionally? Get a publication-ready forest plots with full meta-analysis and subgroup analyses.
Get a Free QuoteLarger squares indicate studies with more weight in the pooled estimate. In random-effects models, weights are more equal across studies because between-study variance is added to each study's variance, reducing the influence of any single large study. The REML estimator produces less biased tau-squared values than DerSimonian-Laird, which can affect weight distribution.
If a study's horizontal confidence interval line crosses the vertical line of no effect (0 for differences, 1 for ratios), that individual study's result is not statistically significant. The same applies to the diamond. Adjusting the confidence level (80%, 90%, 95%, 99%) changes interval width and may affect significance conclusions.
While the confidence interval of the diamond reflects uncertainty about the mean effect, the prediction interval shows the range within which the true effect of a future study is expected to fall. Prediction intervals are always wider than confidence intervals and are especially informative when heterogeneity is substantial, as they reveal whether the effect direction itself is uncertain across settings.
The cumulative meta-analysis tab adds studies one at a time and re-pools at each step, showing how the evidence evolved over time. If the pooled estimate stabilized early and later studies confirm the direction, the evidence is mature. If the estimate is still drifting, more studies may be needed before drawing firm conclusions.
The Galbraith (radial) plot graphs each study's z-score against its precision. Points outside the confidence band are statistical outliers. The Baujat plot graphs each study's contribution to Cochran's Q against its influence on the pooled estimate. Studies in the upper-right quadrant are both heterogeneous and influential, warranting closer scrutiny or sensitivity analysis.
Grouping studies by a categorical moderator (e.g., dose level, population, study design) produces sub-pooled diamonds for each subgroup. The between-group Q-test evaluates whether subgroup differences are statistically significant. A significant Q-between suggests the moderator explains some of the observed heterogeneity.
Leave-one-out analysis removes each study in turn and re-computes the pooled estimate. If the overall effect changes substantially or loses significance when a single study is removed, the meta-analysis is sensitive to that study. Report leave-one-out results to demonstrate the robustness (or fragility) of your conclusions.
The auto-generated R metafor code lets reviewers and co-authors reproduce your exact analysis in R. The methods paragraph generator produces publication-ready text following PRISMA 2020 and Cochrane Handbook conventions, describing the model, estimator, confidence level, and heterogeneity statistics. Both features save time and reduce the risk of reporting errors.
The forest plot is the standard graphical summary for quantitative evidence synthesis, first described by Lewis and Clarke (2001) and formalized in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2023). Every meta-analysis forest plot tool maps three core relationships: individual study effect estimates relate to their precision through inverse-variance weighting, confidence interval lines communicate the uncertainty surrounding each point estimate, and the diamond summary estimate represents the pooled effect across all included studies. This tool goes beyond basic forest plots by offering four visualization tabs: the standard forest plot, a cumulative meta-analysis that shows how the pooled estimate evolves as each study is added, a Galbraith (radial) plot that identifies outlier studies, and a Baujat plot that pinpoints which studies contribute most to heterogeneity and influence the pooled estimate. Software packages such as RevMan (Cochrane's Review Manager) and the metafor package in R are the most widely used platforms for generating forest plots in systematic reviews, and this tool bridges the gap by offering R code generation that produces a complete metafor script for full reproducibility.
When you use a forest plot generator online, the visualization encodes study weight as square size, where larger squares signal studies contributing more to the pooled estimate because they have smaller standard errors. Under the inverse-variance fixed-effect model, weight equals 1/SE². Under random-effects models, between-study variance (τ²) is added to each study's variance, redistributing weight more evenly across studies. This tool supports two tau-squared estimators: the classic DerSimonian-Laird method and the more modern REML (restricted maximum likelihood) estimator, which is generally preferred for its lower bias, especially when the number of studies is small (Veroniki et al., 2016). The Knapp-Hartung adjustment offers a further refinement by using a t-distribution rather than a normal distribution for the pooled confidence interval. This distinction is critical: a meta-analysis forest plot tool must clearly label which model and estimator produced the displayed weights, because the same data can yield materially different pooled estimates and confidence intervals depending on the assumed variance structure.
Publication-ready forest plots require several elements that journals and peer reviewers expect. Study labels (typically author and year) appear on the left axis. Effect estimates with confidence intervals are printed numerically on the right. This tool lets you choose your confidence level from 80%, 90%, 95%, or 99%, adapting the interval width to your field's conventions. A vertical reference line marks the null value: zero for mean differences and standardized mean differences, one for odds ratios and risk ratios. The diamond at the bottom spans the confidence interval of the pooled estimate, with its center at the point estimate. Heterogeneity statistics (I², Cochran's Q with p-value, and τ²) appear below the plot. When prediction intervals are enabled for random-effects models, an additional line below the diamond shows the expected range of true effects in future settings. The font size slider (8-16px) lets you customize text readability for presentations, posters, or journal figures. A free forest plot maker should produce all of these elements without requiring manual annotation, and this tool delivers each one.
Interpreting the pooled effect estimate plot requires attention to both the point estimate and the confidence interval width. A narrow diamond entirely to one side of the null line indicates a statistically significant pooled effect with high precision. A wide diamond crossing the null suggests the pooled evidence is insufficient to conclude a definitive effect. The prediction interval is especially informative when heterogeneity is substantial because it is always wider than the confidence interval and may cross the null even when the confidence interval does not, indicating that the effect direction is uncertain in some settings. The Baujat plot tab graphs each study's contribution to the Q statistic against its influence on the pooled estimate, serving as a complementary influence diagnostic described by Baujat et al. (2002). The Galbraith (radial) plot tab plots standardized effects against precision, with studies outside the confidence band flagged as outliers contributing to heterogeneity (Galbraith, 1988). Subgroup forest plots extend this logic by displaying separate diamonds for predefined subgroups (e.g., by dose, population, or study design), enabling visual comparison of effects across moderators, a technique described in Chapter 10 of the Cochrane Handbook.
Cumulative meta-analysis adds studies one at a time (typically in chronological order) and recalculates the pooled estimate at each step, producing a sequence of pooled effects that reveals how the evidence evolved over time. If the estimate stabilized early and later studies confirmed the direction, the evidence base is mature. If the estimate is still drifting, additional primary studies may be needed before drawing firm conclusions. This technique, formalized by Lau et al. (1992), is a recommended sensitivity analysis in the Cochrane Handbook and is now available as a dedicated tab alongside the main forest plot, the Galbraith plot, and the Baujat plot.
Subgroup analysis extends the standard forest plot by partitioning studies into predefined groups based on a categorical moderator, such as intervention dose, patient population, or study design. Each subgroup receives its own sub-pooled diamond, and the between-group Q-test evaluates whether the subgroup effects differ significantly from one another. A significant Q-between (typically p < 0.10) suggests that the moderator explains a meaningful portion of the observed heterogeneity. This tool includes automatic subgroup detection and computes both within-subgroup pooled estimates and the between-subgroup heterogeneity test. Subgroup analysis should be pre-specified in the review protocol to avoid data-dredging, and the number of subgroups should be kept small relative to the total number of studies to maintain statistical power.
Leave-one-out sensitivity analysis systematically removes each study in turn and recalculates the pooled effect estimate. The result is a series of pooled estimates, one for each iteration, that reveals whether any single study disproportionately drives the overall effect. If the pooled estimate remains stable across all iterations, the meta-analysis is robust. If removing one study causes the effect to change direction, lose statistical significance, or shift substantially in magnitude, that study is influential and warrants closer scrutiny. Common reasons for influence include unusually large sample sizes, extreme effect sizes, or methodological differences. This tool displays the leave-one-out results alongside the main forest plot so you can assess robustness without switching between tools.
The inter-tool pipeline integrates the forest plot generator into a seamless analysis chain. Studies computed in the effect size calculator can be imported directly via the pipeline banner with no re-entry needed. You can also import data by dragging and dropping a CSV or Excel file; the tool uses fuzzy column matching to map headers like "study_name", "Study", or "Author (Year)" to the correct fields automatically. Once your forest plot is complete, send the results downstream to the funnel plot generator for publication bias assessment or to the heterogeneity calculator for deeper analysis, all with a single click from the pipeline workflow bar. Every input is auto-saved to browser storage, so you can close the tab and return later without losing progress. Export options include copying the plot to clipboard as a high-resolution PNG, downloading data as CSV or Excel, saving the plot as PNG or SVG, generating a publication-ready methods paragraph that follows PRISMA 2020 reporting conventions, and copying reproducible R metafor code that lets reviewers and co-authors replicate your exact analysis.
Before generating your systematic review forest plot, ensure all effect estimates are on the same scale. Mixing log odds ratios with raw odds ratios, or combining mean differences measured in different units, produces meaningless pooled estimates. Use our effect size calculator to standardize measures across studies. For reviews pooling prevalence or single-arm proportions, our proportion meta-analysis tool handles Freeman-Tukey, logit, and arcsine transformations designed for this outcome type. Once you have the forest plot, assess between-study consistency: if confidence intervals overlap substantially and I² is below 40%, the evidence is reasonably homogeneous. If I² exceeds 75%, use the Galbraith plot and Baujat plot tabs to identify which studies drive heterogeneity, then consider exploring sources further with our meta-regression data formatter or our leave-one-out sensitivity analysis tool to identify influential studies. To assess whether missing studies may have biased the pooled estimate, generate a funnel plot with Egger's test for publication bias assessment. For small meta-analyses where funnel plots lack power, our Doi plot with LFK index provides a more sensitive alternative for detecting asymmetry.
A forest plot is the standard graphical display for meta-analysis results. It shows individual study effect estimates as squares (sized proportional to study weight) with horizontal lines representing confidence intervals. The pooled summary estimate is displayed as a diamond at the bottom. A vertical line of no effect (at 0 for mean differences or 1 for ratios) provides a visual reference. Forest plots allow readers to assess the magnitude and precision of individual studies, the consistency of effects across studies, and the overall pooled estimate at a glance. This tool supports configurable confidence levels (80%, 90%, 95%, 99%) and optional prediction intervals for random-effects models.
In a fixed-effect meta-analysis (inverse-variance method), each study's weight is proportional to 1/variance (or 1/SE²). Studies with smaller standard errors (more precise estimates) receive more weight. In a random-effects model, an additional between-study variance component (τ²) is added to each study's variance, which reduces the difference in weights between large and small studies. This tool offers two tau-squared estimators: the classic DerSimonian-Laird method and the more modern REML (restricted maximum likelihood) estimator, which is generally preferred for its lower bias. The square size in the forest plot is proportional to the study's percentage weight.
The diamond represents the pooled (summary) effect estimate from the meta-analysis. The center of the diamond is the point estimate, and the left and right tips represent the confidence interval of the pooled effect at your chosen confidence level (80%, 90%, 95%, or 99%). A wider diamond indicates more uncertainty. If the diamond does not cross the line of no effect, the pooled result is statistically significant. When prediction intervals are enabled for random-effects models, an additional line below the diamond shows the range within which the true effect of a future study is expected to fall.
Use a random-effects model when you expect between-study heterogeneity (different populations, interventions, or settings across studies), which is almost always the case in systematic reviews. Fixed-effect models assume all studies estimate the same underlying effect and are appropriate only when studies are functionally identical. Most systematic review guidelines, including Cochrane, recommend random-effects models as the default because they produce more conservative (wider) confidence intervals that account for between-study variability. When using a random-effects model, consider the REML estimator for tau-squared, as it tends to be less biased than DerSimonian-Laird, especially with a small number of studies.
Key heterogeneity statistics include: I² (percentage of variability due to heterogeneity rather than chance: 0-40% low, 30-60% moderate, 50-90% substantial, 75-100% considerable), Cochran's Q (chi-square test for homogeneity, where p < 0.10 suggests significant heterogeneity), and τ² (estimated between-study variance). Visually, overlapping confidence intervals across studies suggest low heterogeneity, while non-overlapping intervals suggest substantial variability. For deeper diagnostics, use the Galbraith (radial) plot tab to spot outlier studies and the Baujat plot tab to identify which studies contribute most to overall heterogeneity.
The diamond represents the pooled (summary) effect estimate from the meta-analysis. Its center indicates the point estimate, and its horizontal tips show the confidence interval at the selected confidence level. A wider diamond indicates less precision. If the diamond does not cross the line of no effect (0 for mean differences, 1 for ratios), the pooled result is statistically significant. When prediction intervals are turned on, a line extending beyond the diamond shows where future true effects are expected to fall.
Each horizontal line represents one study. The square shows the point estimate (its size reflects the study's weight), and the line shows the confidence interval. Studies whose lines cross the vertical line of no effect are not individually significant. The diamond at the bottom is the pooled estimate. Check I² for heterogeneity and the overall p-value for statistical significance. Use the font size slider to adjust text readability, and switch between the forest plot, cumulative meta-analysis, Galbraith plot, and Baujat plot tabs for different perspectives on the data.
Fixed-effect models assume all studies estimate the same true effect and differ only by sampling error. Random-effects models allow the true effect to vary between studies, adding between-study variance (τ²) to the weights. This tool supports two random-effects estimators: DerSimonian-Laird (the classic method) and REML (restricted maximum likelihood, which is less biased). Random-effects models produce wider confidence intervals and give more weight to smaller studies. Enabling prediction intervals on a random-effects plot shows the full expected range of true effects, which is always wider than the confidence interval alone.
A cumulative meta-analysis adds studies one at a time (typically in chronological order) and recalculates the pooled effect estimate at each step. The cumulative meta-analysis tab in this tool shows how the pooled estimate and its confidence interval evolve as each study is added. This visualization reveals whether the evidence stabilized early, whether later studies shifted the conclusion, or whether the pooled estimate is still drifting. Cumulative meta-analysis is a recommended sensitivity technique described in the Cochrane Handbook (Higgins et al., 2023) and is particularly useful for demonstrating temporal trends in a body of evidence.
The Galbraith (radial) plot graphs the standardized effect (z-score) on the y-axis against the inverse of the standard error (precision) on the x-axis. Studies falling outside the confidence band are statistical outliers contributing to heterogeneity. The Baujat plot graphs each study's contribution to the overall Cochran's Q statistic (x-axis) against its influence on the pooled estimate (y-axis). Studies in the upper-right quadrant are both heterogeneous and influential. Together, these diagnostic plots complement I² and Q by identifying exactly which studies drive heterogeneity, helping you decide whether to investigate outliers further or conduct a sensitivity analysis excluding them.
Yes. This tool includes an R code generation feature that produces a complete, ready-to-run script using the metafor package in R. The generated code includes your study data, the selected model (fixed-effect or random-effects with DerSimonian-Laird or REML), confidence level, and all display settings. You can copy the R code to your clipboard with one click and paste it into RStudio or any R environment for full reproducibility. This is especially valuable for peer review, where editors and reviewers may request the underlying analysis code.
The methods paragraph generator produces a publication-ready text block describing the statistical methods used in your meta-analysis. It automatically includes the model type (fixed-effect or random-effects), the tau-squared estimator (DerSimonian-Laird or REML), the confidence level, whether prediction intervals were computed, and the heterogeneity statistics. The paragraph follows reporting conventions from the PRISMA 2020 statement and the Cochrane Handbook. You can copy the paragraph directly into the methods section of your manuscript, saving time and ensuring accurate reporting of your analytical choices.
Assess publication bias visually using our funnel plot and publication bias tool with Egger's test and trim-and-fill analysis. Test the robustness of your pooled estimate with our leave-one-out sensitivity analysis simulator. Calculate individual study effect sizes before plotting with our effect size calculator for SMD, OR, and RR. If you observe high heterogeneity, explore potential moderators with our meta-regression data formatter for R, Stata, or CMA. Pinpoint exactly which studies drive heterogeneity with our GOSH plot generator, visualize meta-regression results with our bubble plot generator, or compare binary event rates across studies with our L'Abbe plot generator. For binary outcome pooling with sparse data, use the Mantel-Haenszel and Peto OR calculator. To evaluate whether your evidence base has reached a conclusive sample size, run a trial sequential analysis. For reviews comparing multiple interventions simultaneously, our network meta-analysis helper supports indirect comparisons and treatment ranking.
Reviewed by
Dr. Sarah Mitchell holds a PhD in Biostatistics from Johns Hopkins Bloomberg School of Public Health and has over 15 years of experience in systematic review methodology and meta-analysis. She has authored or co-authored 40+ peer-reviewed publications in journals including the Journal of Clinical Epidemiology, BMC Medical Research Methodology, and Research Synthesis Methods. A former Cochrane Review Group statistician and current editorial board member of Systematic Reviews, Dr. Mitchell has supervised 200+ evidence synthesis projects across clinical medicine, public health, and social sciences. She reviews all Research Gold tools to ensure statistical accuracy and compliance with Cochrane Handbook and PRISMA 2020 standards.
Our PhD team runs complete meta-analyses: data extraction, effect size computation, forest plots, sensitivity analysis, and a manuscript ready for journal submission. Average turnaround: 2-4 weeks.