How to Read a Forest Plot: A Practical Guide

A forest plot summarizes results from individual studies and, often, a combined analysis called a meta-analysis. It can help you see the direction and size of effects, how precise the estimates are, and whether study results vary. Reading the graphic carefully matters: a plot alone does not establish that an effect is important, reliable, or applicable to a particular situation.

Before you begin, identify the question being studied and the outcome being measured. If the plot comes from a report or review, read its caption and nearby text too. The meaning of an effect estimate depends on the comparison, outcome, and measure used.

What You’ll Need

The forest plot, including its labels, caption, and any footnotes.
The report’s description of the outcome, comparison, effect measure, and confidence level.
A basic understanding that an estimate has uncertainty; specialized statistical software is not necessary for reading the graphic.

Step-by-Step

Read the question and outcome. Find out what groups or conditions are being compared and what outcome is recorded. Check whether a higher outcome is beneficial, harmful, or simply a measurement with no inherently favorable direction. Also note the time frame and population, if given. These details determine what the plotted effects mean.
Identify the effect measure. Look at the horizontal axis and the plot caption for measures such as a risk ratio, odds ratio, hazard ratio, mean difference, or standardized mean difference. For ratio measures, the no-effect value is usually 1. For difference measures, it is usually 0. Do not interpret a ratio as though it were a difference: the scale and its neutral point matter.
Find the no-effect line. This is generally a vertical line through 1 or 0. Values on one side favor one group or direction; values on the other favor the comparison group or opposite direction. Use axis labels, group names, and notes such as “favors treatment” to tell which side means what. Left is not automatically better, and the favorable direction can differ between plots.
Read each study’s estimate. Each row usually represents a study. A square or other marker shows its point estimate: the study’s central result on the chosen scale. Compare the marker with the no-effect line to see the estimated direction, and with the axis labels to judge the approximate size. A marker close to the line indicates an estimate near no effect, not proof that the groups are identical.
Check the confidence interval. A horizontal line extending from the marker usually shows a confidence interval, often a 95% interval. Shorter intervals indicate greater statistical precision; longer ones indicate more uncertainty. If an interval crosses the no-effect line, the result is compatible with no effect at that confidence level. If it stays on one side, it does not cross that threshold, but that alone does not show the result is important or free from bias.
Notice study weights. In many plots, a larger square means that study contributes more weight to the pooled estimate. Weight is usually determined by the analysis method and the amount of information or precision, not simply by how large the square looks or whether a result is favorable. Confirm the legend or methods, since display conventions can vary.
Interpret the pooled result. A diamond often represents the combined estimate: its center is the pooled point estimate, and its width is the confidence interval. Compare the diamond with the no-effect line just as you did for individual studies. A diamond that crosses the line indicates uncertainty about whether the pooled effect differs from no effect under that analysis; it does not establish that the effect is exactly zero.
Assess variation between studies. Look for differences in the study estimates and intervals, and check for an inconsistency measure such as I² or a test of heterogeneity if reported. Variation may reflect differences in participants, methods, settings, or chance. A pooled estimate can conceal meaningful differences, so consider whether combining the studies makes sense and whether the report uses a fixed-effect or random-effects model.
Put the result in context. Read the authors’ interpretation, methods, and limitations. Consider the absolute outcome as well as the relative effect when that information is available. Statistical evidence is not the same as practical importance, and a well-drawn plot cannot correct weak study design, missing data, publication bias, or poor applicability to the question you care about.

Common Mistakes

Treating “not significant” as “no effect.” An interval crossing the neutral line means the data do not rule out no effect at the stated confidence level; it may also include effects that matter in either direction.
Confusing direction with quality. A result favoring one group does not prove that group is better overall. Check outcome definitions, study quality, and possible harms or trade-offs described in the report.
Comparing interval overlap by eye. Whether two confidence intervals overlap is not a reliable standalone test of whether study results differ. Use the review’s formal analysis and interpret it with the study context.
Ignoring the axis scale. Ratio axes are often logarithmic, so equal visual distances may represent multiplicative rather than additive changes. Read the tick labels instead of estimating from spacing alone.
Assuming the diamond settles every question. A pooled result is an average under a particular method. It does not guarantee that the effect applies equally to every group or setting represented.

Tips

Read the figure caption and footnotes before interpreting symbols or abbreviations.
Trace one study row from its name to its estimate, interval, and weight before scanning the whole plot.
Use the axis values to describe both direction and approximate magnitude, rather than saying only “positive” or “negative.”
Separate three questions: which way does the estimate point, how precise is it, and how important might it be?
Check whether the report gives absolute risks or raw outcome values; these can make a relative effect easier to understand.

Troubleshooting

You cannot tell which direction favors which group: Check the axis title, group labels, caption, and any “favors” annotations. If they are missing, consult the report’s methods or results instead of guessing.

The neutral value is unclear: Identify the measure first. Ratios generally use 1 as the no-effect value, while differences generally use 0. If the plot does not name its measure, look for it in the surrounding text; do not infer it from the shape of the axis.

Intervals or markers are hard to distinguish: Enlarge the figure or use a higher-resolution copy. Check the legend for alternate symbols, subgroup markers, or separate interval types. Avoid treating a blurry boundary as an exact value.

Study results point in different directions: Check the subgroup labels and study characteristics, then review heterogeneity information and the analysis model. Do not rely on the pooled estimate alone when important differences remain unexplained.

The pooled estimate seems inconsistent with the individual rows: Check the weights, included studies, and model. The pooled result is not necessarily the simple visual average; ask whether the displayed rows and pooled analysis use the same outcome and set of studies.

Frequently Asked Questions

What does a square mean in a forest plot?

It usually marks an individual study’s point estimate. Its size often represents that study’s weight in the pooled analysis, but check the figure legend to confirm.

What does a diamond mean?

A diamond commonly shows the pooled estimate. Its center marks the combined result, and its width represents the confidence interval.

What does it mean when a confidence interval crosses the vertical line?

For the confidence level shown, the interval includes the no-effect value. The result does not clearly distinguish an effect from no effect at that threshold, but this is not proof of no effect.

Is a narrow confidence interval always better?

A narrow interval indicates greater precision, but it does not guarantee that the estimate is unbiased, relevant, or important. Study design and applicability still matter.

Can a forest plot show which treatment is best?

Not by itself. It summarizes specific outcomes and comparisons; deciding what a result means also requires evidence quality, absolute effects, trade-offs, and the context described in the source.

Further reading: Cochrane. Details can change, so check the official page for anything that matters to you. Spotted a mistake? Tell us. How we write an answer.