Sep 14, 2026
How to Read a Funnel Plot: Asymmetry, Small-Study Effects, and Publication Bias
By JournalLabs Research Team
Introduction
A funnel plot can help researchers examine whether effect estimates differ systematically between smaller and larger studies in a meta-analysis. Each dot represents a study, while its position reflects the study's estimated effect and a measure of its size or precision.
The plot is frequently described as a way to detect publication bias. That description is incomplete. An asymmetric funnel plot shows a pattern known as small-study effects, but selective publication is only one possible explanation. Heterogeneity, methodological differences, statistical artifacts, and chance can produce similar shapes.
This guide explains how to read the axes and distribution of points, investigate asymmetry, interpret statistical tests, and report the result without treating the plot as proof of missing studies.
What Does a Funnel Plot Show?
A funnel plot is a scatter plot of study effect estimates against a measure of study size or precision. It is usually created for one meta-analysis containing studies that address the same outcome and comparison.
The common elements are:
| Element | Typical meaning | What to check |
|---|---|---|
| Dot | One study estimate | Confirm what each point represents |
| Horizontal axis | Effect size | Identify the measure and direction |
| Vertical axis | Standard error, precision, or sample size | Check whether the scale is reversed |
| Center line | Often the pooled effect | Do not assume it is the no-effect line |
| Funnel limits | Expected sampling variation | They are not study confidence intervals |
When standard error is on the vertical axis, more precise studies usually appear near the top because smaller standard errors are plotted higher. Less precise studies appear near the bottom and scatter more widely through sampling variation. This produces the expected inverted-funnel shape.
Some plots use precision or sample size instead. Their vertical direction and geometry may differ. Always read the axis labels instead of assuming that every funnel plot follows the same convention.
Unlike a forest plot, a funnel plot is not primarily designed to show each study's confidence interval or the pooled effect. It examines the relationship between estimated effects and study size or precision.
How to Read a Funnel Plot Step by Step
1. Identify the Meta-Analysis Being Plotted
Start with the outcome, comparison, effect measure, population, and set of included studies. A review may contain a separate funnel plot for each outcome or subgroup.
Do not combine points from clinically unrelated analyses simply to increase the number of studies. Genuine differences in populations, interventions, or follow-up periods can create asymmetry that has little to do with missing evidence.
2. Read the Horizontal Axis
The horizontal axis shows the effect estimate, such as a risk ratio, odds ratio, mean difference, or standardized mean difference. Ratio measures are commonly displayed on a logarithmic scale.
Determine which direction favors the intervention and where the no-effect value lies. For ratios, no effect is usually 1; for differences, it is usually 0. The vertical center line of the funnel may instead mark the pooled estimate, so it should not automatically be interpreted as no effect.
3. Read the Vertical Axis
The vertical axis represents a measure related to study size or precision. Standard error is common: large, more precise studies appear toward the top, while small, less precise studies appear toward the bottom.
If the plot uses precision, higher values may appear at the top. If it uses sample size, interpretation depends on how the scale is drawn. The chosen axis matters because it affects the expected funnel shape and the appearance of asymmetry.
4. Examine the Top of the Plot
The most precise studies should generally cluster more closely because their estimates have less sampling variation. Observe where these studies sit and whether they suggest a stable central effect.
Do not automatically declare the largest study correct. It may still have design limitations or address a different setting. However, a consistent difference between the larger studies at the top and smaller studies below is an important pattern to investigate.
5. Examine the Spread at the Bottom
Smaller studies should scatter more widely on both sides of the center. Ask whether the lower part contains points on both sides or whether one region appears sparsely populated.
A missing-looking lower corner may raise concern, but the plot does not reveal whether studies truly exist in that space. Visual judgment is subjective, particularly when there are few points, several studies have similar precision, or one extreme study changes the apparent shape.
6. Compare the Two Sides Without Overinterpreting Them
A roughly balanced shape means the scatter appears similar around the center. It does not prove that every relevant study was published or found. Selective reporting can occur without producing visible asymmetry.
An unbalanced shape indicates that effect estimates vary with study size or precision more than expected visually. This is evidence to investigate possible small-study effects-not a diagnosis of publication bias.
What Causes Funnel Plot Asymmetry?
Non-Reporting or Publication Bias
Studies or results may be less likely to become available when their findings are statistically non-significant, unfavorable, or less striking. If missing results are concentrated among small studies on one side of the plot, the observed evidence may look asymmetric.
Publication bias is only part of the broader problem of missing evidence. Selective outcome reporting, delayed publication, inaccessible results, and selective dissemination can also affect what enters a meta-analysis.
Methodological Differences Between Small and Large Studies
Smaller studies may use weaker allocation methods, less complete follow-up, selective analyses, or other practices that exaggerate effect estimates. If such limitations occur more often in small studies, their estimates can differ systematically from those of larger studies.
Study size itself does not prove quality. The explanation must be examined using risk-of-bias assessments and study characteristics.
Genuine Heterogeneity
Small and large studies may evaluate different populations, doses, settings, or implementation methods. For example, early small studies might enroll higher-risk participants who experience a larger absolute or relative benefit.
This can create a real relationship between study size and effect. Investigate whether the pattern changes across clinically meaningful subgroups rather than assuming the smaller estimates are distorted.
Statistical Artifacts
For some effect measures, the estimated effect and its standard error are mathematically related. Sparse events, standardized mean differences, inappropriate scales, or the selected test can create apparent asymmetry even without selective publication.
Chance
With a limited number of studies, random scatter may look unbalanced. The human eye is especially likely to find a pattern when only a few points define the funnel.
Funnel Plot Example: Interpreting an Asymmetric Pattern
Consider a fictional meta-analysis of 14 trials comparing an intervention with control for an undesirable outcome. The horizontal axis displays log risk ratios, and the vertical axis displays standard error with more precise trials at the top.
Three large trials cluster near a risk ratio of 0.95. Smaller trials scatter more widely, but most of them report stronger benefits, with relatively few small studies showing no benefit or harm. The lower-right portion of the funnel therefore looks sparse.
A careful interpretation would be:
The funnel plot suggests small-study effects because smaller trials tend to report more favorable estimates than larger trials. Selective non-publication of small, unfavorable studies is one possible explanation, but differences in populations, methods, and sampling variation should also be investigated.
It would be inaccurate to write that the plot "proves publication bias." The next checks should include:
- Comparing the clinical and methodological characteristics of small and large trials.
- Reviewing study protocols and trial registries for completed studies or missing outcomes.
- Examining risk of bias across studies.
- Using an appropriate statistical test if the analysis meets its conditions.
- Testing whether conclusions change under reasonable sensitivity analyses.
If the supposedly missing region contains studies that would have been statistically non-significant and unfavorable, a contour-enhanced funnel plot may make selective non-reporting more plausible. It still cannot identify missing studies by itself.
Statistical Tests for Funnel Plot Asymmetry
What Does Egger's Test Evaluate?
Egger's test is a regression-based test for an association between study effects and their standard errors. A small P-value can indicate more asymmetry than expected by chance under the test's assumptions.
It does not directly test whether unpublished studies exist. A statistically significant result supports the presence of small-study effects, which may have several causes. A non-significant result does not rule them out because these tests often have low power.
When Should a Test Be Used?
A common rule of thumb is to test funnel plot asymmetry only when the meta-analysis contains at least 10 studies. With fewer studies, the test is unlikely to distinguish real asymmetry from chance reliably.
Even with 10 or more studies, testing may be uninformative when study sizes are very similar. The method must also fit the effect measure and data. The original Egger test is not appropriate in every situation, including some analyses of odds ratios or standardized mean differences where effect estimates can be correlated with their standard errors.
How Should Tests and Sensitivity Analyses Be Combined?
Interpret the statistical result alongside the plot, heterogeneity, study characteristics, and evidence about missing results. No single P-value settles the explanation.
Useful sensitivity checks may compare fixed-effect and random-effects estimates, examine the influence of smaller studies, or apply selection models under explicit assumptions. Trim-and-fill may estimate how a pooled result changes under an assumed missing-study pattern, but it neither proves which studies are missing nor automatically corrects publication bias.
Common Funnel Plot Interpretation Mistakes
Treating Asymmetry as Proof of Publication Bias
Asymmetry displays small-study effects. Publication bias is one possible cause, not the only conclusion.
Using a Funnel Plot With Too Few Studies
A plot with only a handful of points has little ability to reveal a stable pattern. Avoid strong visual or statistical claims when the evidence base is small.
Ignoring the Axis and Effect Measure
The apparent direction of missing studies depends on the scale, effect measure, outcome, and comparison. Read the labels before deciding which side is favorable.
Mixing Incomparable Studies
Combining different outcomes or clinically distinct subgroups can manufacture asymmetry through heterogeneity. Plot the studies belonging to the same defensible meta-analysis.
Calling a Symmetric Plot Evidence of No Bias
Symmetry does not rule out missing evidence. Bias may affect both sides, involve outcomes rather than whole studies, or remain invisible because the analysis lacks power.
How JournalLabs Supports Research Paper Interpretation
JournalLabs' AI research paper summarizer helps researchers identify and organize a paper's methods, reported findings, and limitations before returning to the original evidence. This context can clarify how authors created and interpreted a funnel plot within a systematic review or meta-analysis.
The feature does not diagnose publication bias or replace statistical analysis. Researchers should inspect the original plot, verify the extracted information, and interpret asymmetry alongside the paper's methods, study characteristics, risk-of-bias assessments, and sensitivity analyses.
Frequently Asked Questions
Does an asymmetric funnel plot prove publication bias?
No. It suggests small-study effects. Selective publication is one possible explanation, but heterogeneity, study limitations, statistical artifacts, and chance can also produce asymmetry.
How many studies are needed for a funnel plot?
There is no number that makes visual interpretation certain. For formal asymmetry tests, at least 10 studies is a common rule of thumb. Even then, the range of study sizes and the suitability of the test matter.
What does each dot represent?
Each dot normally represents one study estimate included in a particular meta-analysis. Check the figure legend because some plots may use multiple estimates, subgroups, or alternative units.
What is the difference between a funnel plot and a forest plot?
A forest plot displays study estimates, confidence intervals, weights, and often a pooled result. A funnel plot displays effect estimates against study size or precision to investigate small-study effects.
Can a symmetric funnel plot rule out publication bias?
No. The plot may have too few studies, missing evidence may not create asymmetry, or different biases may offset each other. Symmetry is not proof that the evidence set is complete.
Conclusion
To read a funnel plot, identify the effect measure and precision axis, examine how estimates spread from larger to smaller studies, and look for a relationship between study size and effect.
Describe asymmetry as possible evidence of small-study effects. Then investigate publication and reporting bias alongside heterogeneity, methodological differences, statistical artifacts, and chance. A responsible interpretation combines the plot with study-level evidence and suitable sensitivity analyses rather than relying on shape alone.
Interpret the Pattern, Not Just the Shape
Use JournalLabs to identify a paper's meta-analysis methods, reported findings, and limitations, then return to the original funnel plot before drawing conclusions about small-study effects or missing evidence.