A funnel plot is a scatter plot used in meta-analysis that shows whether the collection of studies being combined might be missing data, most commonly due to publication bias. Each dot represents one study, and the overall shape of the scatter tells you whether the evidence base looks complete or skewed. Reading one takes only a few minutes once you understand what the axes mean and what patterns to look for.
What the Axes Represent
The horizontal axis (x-axis) shows each study’s effect size, which is the result that study found. Depending on the type of research, this might be an odds ratio, a risk ratio, or a mean difference. Studies landing further from the center line found larger or smaller effects.
The vertical axis (y-axis) shows a measure of each study’s precision, most commonly the standard error. Smaller standard errors mean more precise estimates, and these appear at the top of the plot. Larger, less precise studies sit near the bottom. Some funnel plots flip this convention or use sample size instead, but the logic is the same: one direction means “more reliable” and the other means “less reliable.”
Why It Looks Like a Funnel
Large, precise studies tend to produce results that cluster tightly around the true effect. They appear as dots near the top of the plot, close together. Smaller studies have more random variation in their results, so their dots spread out widely near the bottom. This natural pattern creates an inverted triangle, or funnel shape, widening as you move down the plot. The “funnel” isn’t drawn by the software in most cases. It’s the expected shape the dots should form if everything is working normally.
Reading a Symmetrical Plot
A symmetrical funnel plot is the ideal scenario. It means the studies scatter evenly on both sides of the average effect, with roughly equal numbers of dots to the left and right at every level of precision. This symmetry suggests the meta-analysis has captured a representative sample of the research. Small studies that found smaller effects and small studies that found larger effects are both present. No obvious chunk of evidence is missing.
Symmetry alone doesn’t guarantee the absence of publication bias, but it’s a reassuring sign. Think of it as passing a visual sanity check before looking deeper.
What Asymmetry Tells You
When dots cluster more heavily on one side, especially among the smaller studies near the bottom of the plot, the funnel looks lopsided. This is the pattern that raises concern about publication bias, the tendency for studies with statistically significant or “interesting” results to get published while studies with null findings sit in file drawers.
A gap in one corner of the funnel is the classic red flag. If you see plenty of small studies showing large positive effects but very few small studies showing small or negative effects, it suggests those less exciting results may exist but were never published. The meta-analysis is then working with an incomplete picture, and its overall estimate could be inflated.
Publication bias isn’t the only explanation for asymmetry, though. Other causes include genuine differences between studies (heterogeneity), differences in study quality where smaller trials used weaker methods, or even chance. A lopsided funnel plot is a starting point for investigation, not a definitive diagnosis.
Contour-Enhanced Funnel Plots
A more advanced version adds shaded regions representing levels of statistical significance, typically marking the boundaries for p-values below 0.01, 0.05, and 0.10. These contour lines help you figure out why the plot is asymmetric.
If the “missing” studies would fall in areas of statistical nonsignificance, that supports the idea of publication bias. Researchers or journals may have suppressed results that didn’t reach the significance threshold. If the gap is instead in areas of higher statistical significance, the asymmetry is more likely caused by something else entirely, such as differences in study quality or true variation in effects across populations. Contour-enhanced plots give you a second layer of information that a standard funnel plot can’t provide on its own.
Statistical Tests for Asymmetry
Eyeballing a funnel plot is subjective, so formal statistical tests exist to quantify what you’re seeing. The two most common are Egger’s regression test and Begg’s rank correlation test.
Egger’s test fits a regression line through the study results. If no publication bias exists, the line’s intercept should be zero. A significant intercept suggests the smaller studies are systematically pulling in one direction. Because these tests generally have low statistical power, the threshold for significance is typically set at 10% (p < 0.10) rather than the usual 5%.
Begg’s test uses a different approach based on ranking the studies rather than fitting a regression line. It tends to be less sensitive than Egger’s test under standard conditions but performs better when study results have unusual distributions with extreme outliers or heavy skew. In practice, many researchers report both tests together.
How Many Studies You Need
Funnel plots become unreliable when a meta-analysis includes too few studies. The Cochrane Handbook, the standard reference for systematic reviews, recommends using funnel plot tests only when there are at least 10 studies. With fewer than that, the plot simply doesn’t have enough data points to form a meaningful shape, and statistical tests for asymmetry lack the power to detect real bias. If you’re looking at a funnel plot built from six or seven studies, treat any interpretation with caution.
The Trim-and-Fill Method
When a funnel plot does suggest missing studies, researchers can try to estimate how those missing results would have changed the overall conclusion. The trim-and-fill method does this in two steps. First, it temporarily removes (trims) the small asymmetric studies from one side of the funnel to create a symmetric shape, then recalculates the overall effect using only the remaining, larger studies. Second, it adds back (fills) the trimmed studies along with hypothetical mirror-image counterparts on the opposite side, producing an adjusted estimate that accounts for the suspected missing data.
This adjusted estimate gives you a sense of how much the overall result might shift if the missing studies were included. It’s a useful sensitivity check, but it relies on the assumption that the asymmetry is caused by publication bias specifically. If the real cause is something else, the “correction” could actually introduce error rather than fix it.
Putting It All Together
When you encounter a funnel plot in a published review, work through it in order. First, check the axes to understand what’s being plotted. Then look at the overall shape: does it resemble a symmetric inverted funnel, or is it visibly lopsided? Note where the gaps are, particularly among the smaller studies at the bottom. If contour lines are present, check whether the missing area falls in zones of nonsignificance. Finally, look for any accompanying statistical tests and trim-and-fill analyses in the text.
No single element gives you a definitive answer. A symmetric plot with a nonsignificant Egger’s test is reassuring. An asymmetric plot with a significant test and missing studies in nonsignificant zones is a strong signal that publication bias may be inflating the result. Everything in between requires judgment, and the funnel plot is one piece of that puzzle rather than the whole picture.