Forest plot

A forest plot is the graphical display used in a meta-analysis to show individual study effect estimates alongside the pooled (combined) summary estimate. It is the standard figure format specified or exemplified in the PRISMA 2020 reporting guideline and in the Cochrane Handbook’s chapter on meta-analysis, and it is normally what a reader looks to first when assessing a meta-analysis‘s results.

What makes a chart a forest plot

A forest plot has a consistent structure regardless of the software used to generate it (RevMan, R’s meta/metafor packages, Stata, and similar tools all produce the same layout):

• One row per included study. Each row shows the study’s point estimate as a square (or sometimes a diamond or circle) positioned on the effect-size axis, with a horizontal line extending through it marking the study’s confidence interval — conventionally 95%.
• Square size reflects study weight. Larger squares indicate studies that contribute more weight to the pooled estimate, typically because they have larger sample sizes or narrower confidence intervals (greater precision).
• A vertical reference line at the null value. This is placed at 0 for difference-based effect measures (mean difference, risk difference) or at 1 for ratio-based measures (odds ratio, risk ratio, hazard ratio). If a study’s confidence-interval line crosses this reference line, that individual study’s result is not statistically significant on its own.
• A summary diamond at the bottom. This represents the pooled effect estimate produced by combining all included studies statistically. The diamond’s horizontal center marks the pooled point estimate; its left and right tips mark the pooled estimate’s confidence interval. If the diamond does not cross the null-value reference line, the pooled result is statistically significant.
• A numeric summary column alongside the plot, typically listing each study’s effect estimate, confidence interval, and percentage weight, plus the pooled statistics and a heterogeneity statistic (commonly I²).

Why forest plots matter

A forest plot lets a reader assess two things at a glance that a table of numbers does not make as visible: whether the pooled effect is being driven consistently across studies or by one or two outliers, and how much the individual study estimates disagree with each other (heterogeneity). Widely scattered point estimates with little overlap between confidence intervals is a visual signal of heterogeneity, which is normally also reported as a formal statistic (I²) alongside the plot. Per the Cochrane Handbook, I² interpretation bands deliberately overlap and should not be read as a mechanical cutoff — the plot itself, showing where the disagreement actually sits, is part of how a reader is meant to judge it rather than relying on the summary statistic alone.

Forest plots also make it possible to see immediately whether a particular subgroup or a particular study is pulling the pooled estimate in one direction, which is why sensitivity analyses and subgroup analyses in a meta-analysis are usually presented as their own forest plots.

Worked example

Consider a meta-analysis of eight randomized controlled trials comparing a treatment to placebo, pooled using a random-effects model, reporting odds ratios (OR):

• Each of the eight trials appears as one row, with a square at its OR and a horizontal line spanning its 95% confidence interval.
• The vertical reference (null) line sits at OR = 1.0, since an odds ratio of 1 means no difference between treatment and placebo.
• Trials with wide confidence intervals that cross OR = 1.0 are not statistically significant individually; a trial whose entire confidence interval sits to the left of 1.0 favors treatment on its own.
• The diamond at the bottom shows the pooled OR (for example, 0.78) with its own 95% confidence interval (for example, 0.68–0.89). Because the diamond does not cross OR = 1.0, the pooled result is statistically significant in favor of treatment, even though not every individual trial was.
• An I² statistic reported alongside (for example, I² = 45%) indicates the degree of heterogeneity among the eight trials’ individual estimates.

What is not a forest plot

A generic scatter plot or bar chart of effect sizes across studies, without confidence-interval lines and a pooled summary diamond, is not a forest plot — it is missing the specific elements (per-study CIs, weighting reflected in marker size, and a combined pooled estimate) that define the format. A funnel plot is a related but distinct chart used in meta-analysis: it plots each study’s effect estimate against a measure of precision (such as standard error) to visually assess asymmetry associated with small-study effects and possible publication bias, rather than displaying individual confidence intervals row by row or a pooled diamond.

Related CASRAI content

• Meta-analysis — the statistical procedure a forest plot visualizes.
• Systematic review — the broader review process a meta-analysis (and its forest plot) is often embedded within.
• Systematic Review vs. Meta-Analysis — comparison page.
• PRISMA and Systematic Review Methodology — guide.
• PRISMA 2020 — the reporting guideline that specifies systematic-review and meta-analysis reporting, including result figures such as forest plots.

References

• Cochrane Handbook for Systematic Reviews of Interventions, Chapter 10 (Analysing data and undertaking meta-analyses) — training.cochrane.org/handbook
• Page MJ, McKenzie JE, Bossuyt PM, et al. “The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.” BMJ. 2021;372:n71.