Sep 8, 2026
How to Read a Forest Plot: Effect Sizes, Confidence Intervals, and the Overall Result
By JournalLabs Research Team
Introduction
A forest plot can condense the findings of several studies into one figure. Squares show individual estimates, horizontal lines show uncertainty, and a diamond often summarizes the combined result. The display is compact, but a quick glance can lead to the wrong conclusion.
To read a forest plot, identify the outcome and effect measure first. Then locate the line of no effect, examine each estimate and its confidence interval, check study weights, and interpret the pooled result alongside differences between studies.
Neither a large square nor a narrow diamond proves that the evidence is trustworthy. This guide explains the visual elements, walks through a worked example, and shows how to describe the findings without confusing statistical significance, practical importance, and certainty.
What Does a Forest Plot Show?
A forest plot displays effect estimates and their uncertainty on a shared scale. It commonly appears in a meta-analysis, where results from multiple studies are combined. Similar plots can also display subgroup estimates or different analyses within one study.
Always check what each row represents. It might be an independent study, a subgroup, a model, or a sensitivity analysis. A forest plot does not automatically mean that every row can be treated as a separate source of evidence.
In a typical meta-analysis figure, the main elements are:
| Element | Meaning | Reading check |
|---|---|---|
| Square or marker | Individual effect estimate | Position shows the estimated effect |
| Horizontal line | Usually a 95% confidence interval | Width indicates precision |
| Square area | Often proportional to study weight | Not a quality rating |
| Vertical reference line | No effect or no difference | Value depends on the measure |
| Diamond | Pooled estimate and its confidence interval | Check center and both tips |
Some plots omit weights or pooled estimates. Others contain multiple diamonds for different outcomes or subgroups. Read the title, legend, and column labels before interpreting the symbols.
How to Read a Forest Plot Step by Step
1. Identify the Outcome and Comparison
Start with what was measured, which groups were compared, and the follow-up period. A plot of symptom improvement answers a different question from a plot of adverse events, even if both use the same effect measure.
Check which group is the numerator or reference category. An estimate below the no-effect value is not automatically beneficial. Lower values may favor the intervention for an undesirable event but favor the comparator when the outcome is recovery.
2. Find the Effect Measure and Line of No Effect
For risk ratios, odds ratios, and hazard ratios, the no-effect value is 1. A ratio of 1 indicates equal risk, odds, or hazard, respectively, for the comparison being reported.
For mean differences, standardized mean differences, and risk differences, the no-effect value is 0. Zero indicates no difference on that measure.
For a mean difference, identify the measurement unit and subtraction order. A difference of five points means little without knowing the scale and whether higher scores indicate improvement. A standardized mean difference expresses the difference in standard-deviation units, not the original measurement units.
Do not interpret an odds ratio as if it were a risk ratio. They describe different quantities, and the distinction matters when an outcome is common. Identify the measure before translating the number into ordinary language.
3. Read the Axis Scale
Ratio measures are commonly displayed on a logarithmic axis. Equal distances then represent equal multiplicative changes: 0.5 and 2 are equally far from 1 on a log scale.
This is why tick labels may look uneven numerically. A ratio cannot take the value zero on a logarithmic axis. Difference measures are often shown on a linear scale instead.
If a confidence interval extends beyond the displayed range, an arrow may mark the clipped end. Consult the numerical interval rather than treating the visible segment as the complete uncertainty range.
Also check the axis limits when comparing two plots. The same numerical interval can look longer or shorter when the displayed range changes. Compare the reported estimates and limits, not only the physical lengths of the lines.
4. Examine Individual Confidence Intervals
A horizontal line usually represents a 95% confidence interval around an estimate. A wider interval indicates less precision under the analysis assumptions; a narrower interval indicates greater precision. It is not the range containing 95% of individual participants' outcomes.
For a conventional two-sided test at the 5% level, a corresponding 95% confidence interval that excludes the no-effect value generally indicates statistical significance. Check the paper's reported methods and numbers, especially when an endpoint is rounded to 1.00 or 0.00.
An interval crossing the reference line does not prove that the effect is zero. It means the interval includes no difference alongside other possible effect sizes.
5. Check Study Weights
In many plots, square area reflects how much a study contributes to the pooled estimate. Under fixed-effect inverse-variance weighting, a more precise estimate receives greater weight.
Weight is not simply participant count. Event frequency, outcome variability, study design, and the statistical method can affect precision and contribution. A random-effects model also incorporates estimated between-study variance and can change the relative weights.
A large square does not certify better methods or lower bias. Statistical weight and evidence quality are separate considerations.
6. Interpret the Diamond
The diamond's center marks the pooled estimate, and its horizontal tips usually mark the confidence interval. Check whether the entire diamond lies on one side of the no-effect line.
The pooled result is a model-based summary, not a vote among significant studies. It can be statistically significant even when several individual studies have intervals crossing the reference line.
Before using it, read the model label and check whether the combined studies address a sufficiently comparable question.
Forest Plot Example: Reading Four Studies
The following figure uses four fictional studies to illustrate interpretation. It compares intervention versus control for an undesirable event. The studies are assumed to address a common outcome and follow-up period for teaching purposes; they are not clinical evidence.
Figure 1. Teaching example using fixed-effect inverse-variance weighting of log risk ratios. Numbers are rounded for display.
Read the Individual Studies
Study A has a risk ratio of 0.67 and a 95% confidence interval of 0.49 to 0.92. Its estimate favors the intervention for this undesirable outcome, and its interval excludes 1.
Study B has a risk ratio of 0.80, with an interval of 0.63 to 1.02. The point estimate also favors the intervention, but the interval includes 1. It receives the largest weight, 42.2%, because its estimate is the most precise under the specified model.
Study C has a risk ratio of 0.89, with an interval of 0.62 to 1.26.
Study D has a risk ratio of 1.05, with an interval of 0.71 to 1.56. Both intervals include 1. Study D's point estimate is above 1, but its uncertainty does not establish a clear increase in risk.
Read the Combined Result
The pooled risk ratio is 0.82, with a 95% confidence interval of 0.70 to 0.95. The diamond is entirely below 1, indicating a statistically significant result under this illustrative analysis.
The point estimate corresponds to approximately 18% lower relative risk, not an 18-percentage-point reduction. To illustrate the difference, if baseline risk were 10%, applying a risk ratio of 0.82 would imply approximately 8.2% risk, a reduction of 1.8 percentage points. That baseline risk is hypothetical, not supplied by the plot.
A suitable summary is:
In this teaching example, the pooled estimate favors the intervention, although three individual study intervals include no difference. The plot alone does not establish practical importance, low risk of bias, or applicability to other populations.
What the Overall Result Does Not Tell You
Heterogeneity Requires More Than a Visual Check
Studies can differ in populations, interventions, measurements, follow-up, and design. Inspect the spread of estimates, then read the reported heterogeneity statistics and explanations.
I-squared describes the estimated proportion of variability in effect estimates attributable to heterogeneity rather than sampling error. It is not the percentage of studies that disagree or the probability that the conclusion is wrong.
The Cochrane Handbook cautions against interpreting heterogeneity through simple thresholds alone, especially with few studies. A low I-squared value does not establish that all studies are clinically equivalent or that important differences are absent.
The Model Changes the Interpretation
A conventional fixed-effect model assumes that studies estimate a common underlying effect. A random-effects model allows underlying effects to differ and summarizes their average within the assumed distribution.
The diamond's confidence interval is not a prediction interval. In a random-effects analysis, a prediction interval can help describe the spread of underlying effects across comparable settings, subject to substantial uncertainty when evidence is limited.
The teaching figure above uses a fixed-effect model for illustration, not because that model is automatically appropriate for every review.
Statistical Significance Does Not Establish Importance
An estimate may exclude no effect yet be too small to matter in practice. Consider the absolute effect, baseline risk, benefits and harms, and any relevant threshold for a meaningful difference.
Likewise, a precise pooled result can still be affected by biased studies or missing evidence. The forest plot summarizes the included estimates; it cannot show all unpublished or unavailable results.
Common Forest Plot Interpretation Mistakes
Assuming the Left Side Always Means Better
Direction depends on the outcome and comparison. Read the axis labels and outcome definition before using phrases such as "favors treatment."
Counting Significant Studies
Do not conclude that an intervention works because more studies are significant, or that it fails because most are not. Examine effect sizes, intervals, weights, and the pooled model.
Calling Different Subgroup P-Values a Difference in Effect
One subgroup being significant and another not significant does not establish that their effects differ. Look for an appropriate subgroup comparison or interaction test rather than comparing significance labels.
Treating the Diamond as the Whole Review
A diamond cannot replace assessment of study quality, missing evidence, or population relevance. A narrow pooled interval does not remove limitations in the evidence used to calculate it.
How JournalLabs Supports Evidence-Aware Paper Reading
JournalLabs' AI research paper summarizer helps researchers organize a paper's methods, findings, and limitations before checking the original evidence.
For a paper containing a forest plot, use that context to identify the outcome, comparison, reported model, and limitations that shape interpretation. Then inspect the figure, legend, and numerical results directly. Do not assume that a generated summary has interpreted every plotted value correctly or that the tool has independently validated the meta-analysis.
Frequently Asked Questions
What does the diamond mean in a forest plot?
It usually represents a pooled estimate. Its center marks the estimate, while its left and right tips show the confidence interval. Check the legend because display conventions can vary.
What if a confidence interval crosses the no-effect line?
The interval includes no difference. Under the corresponding conventional test, the result is not statistically significant, but this does not prove the absence of an effect.
Why do some forest plots use 0 and others use 1?
Ratio measures have a no-effect value of 1. Difference measures have a no-effect value of 0. Identify the reported measure rather than memorizing one reference value for every plot.
Does a bigger square mean a better study?
No. It commonly indicates greater statistical weight in that analysis. Risk of bias and methodological quality need separate assessment.
Can a forest plot appear without a pooled diamond?
Yes. Authors may display individual estimates without pooling them, particularly when combining results would be inappropriate. Similar plots can also present multiple estimates from one study.
Conclusion
Reading a forest plot starts with the research question, not the diamond. Identify the outcome, comparison, effect measure, and scale before interpreting estimates, confidence intervals, weights, and the pooled result.
Then consider what the figure leaves unresolved: heterogeneity, model assumptions, bias, practical importance, and applicability. A useful interpretation explains the direction, magnitude, and uncertainty of the evidence without reducing the entire review to "significant" or "not significant."
Connect the Plot to the Paper
Use JournalLabs to organize the study context, reported findings, and limitations. Return to the original forest plot and its supporting methods before including the result in research notes or a literature review.