On this page
Interactive reading practice
Inspect one forest plot two ways
All values below are constructed for this non-clinical lesson. They are not observations about real learners, products or outcomes.
| Study | Difference, minutes | 95% confidence interval | Assigned weight |
|---|---|---|---|
| Study A | -4.0 | -8.0 to 0.0 | 25.9% |
| Study B | -1.0 | -4.0 to +2.0 | 46.0% |
| Study C | +2.0 | -4.0 to +8.0 | 11.5% |
| Study D | -2.0 | -7.0 to +3.0 | 16.6% |
| Fixed-effect summary | -1.6 | -3.6 to +0.4 | 100% |
For this teaching example only, each input interval is treated as a conventional normal-approximation 95% interval; inverse-variance weights and the fixed-effect summary are calculated from those constructed values. This is not a recommendation to combine real studies.
| Study | Difference, minutes | 95% confidence interval |
|---|---|---|
| Study A | -4.0 | -8.0 to 0.0 |
| Study B | -1.0 | -4.0 to +2.0 |
| Study C | +2.0 | -4.0 to +8.0 |
| Study D | -2.0 | -7.0 to +3.0 |
What the constructed example shows
- Zero is the null value for the mean difference plotted here.
- The square area reflects the inverse-variance weight assigned in the pooled view; it is not a direct display of sample size.
- The summary interval crosses zero, so both directions remain compatible with this constructed estimate and its stated interval.
- The individual-studies view intentionally omits the pooled estimate and weights.
Read the outcome, comparison and null value first
A forest plot is not self-explanatory. Read the outcome, units, comparison direction and effect measure before interpreting a mark. In this constructed example, the measure is a mean difference in minutes: revised workflow minus current workflow. A negative value means fewer minutes with the revised workflow; a positive value means more minutes. Zero is the null value for this difference measure.
A ratio measure uses a different scale: its null is 1, not 0. The vertical reference line therefore has to be read with the axis label. Do not borrow the direction or null value from another forest plot. “Left” is not automatically better; it only has a meaning after you read which group was subtracted from which.
Separate the study estimate, interval and weight
For an individual study, the square marks its point estimate and the horizontal line shows its confidence interval. The square’s area represents the weight assigned to that study in this constructed inverse-variance synthesis. It does not simply report the number of participants. The interval communicates uncertainty around the estimate; it is not the range containing most individuals’ outcomes.
Study B’s interval runs from −4 to +2 minutes, so it includes zero. That keeps no difference among the values compatible with this interval under the analysis assumptions; it does not prove identical outcomes. Inspect the estimate, interval, measurement and study design together.
Supporting reading: [1]
Read a pooled diamond only when one is shown
The first view shows a fixed-effect summary calculated from the four constructed study estimates. The diamond’s center is the pooled estimate (about -1.6 minutes); its left and right points mark the confidence interval (about -3.6 to 0.4 minutes). That interval reaches across zero, so this fictional example does not establish a direction of difference. The values are teaching inputs, not a study result.
A forest plot can also display separate study results without a pooled diamond. The second view presents the same individual estimates without a summary row or meta-analysis weights. Whether combining studies is appropriate depends on the review question, outcome, study methods and assumptions; the presence of a plot does not establish that pooling occurred or should occur.
Do not use one visual cue as a heterogeneity test
Poor overlap among study confidence intervals can be a visual sign of statistical heterogeneity, but the plot alone is not a complete assessment. Cochrane describes a formal test and measures such as I², each with limits; a small number of studies makes heterogeneity harder to assess reliably. Different populations, measures, bias and reporting also affect how a synthesis should be interpreted.
A confidence interval around a pooled mean and a prediction interval are not interchangeable. The former describes uncertainty in the summary estimate; a prediction interval aims to describe the spread of underlying effects in a similar new study, and can be unreliable when few studies are available. Check which interval the review actually reports.
Write down what the plot supports and what remains unknown
A useful reading record names the outcome and unit, subtraction direction, effect estimate, interval, model and any displayed heterogeneity information. Then state what the plot cannot answer alone: whether the studies are unbiased, whether the outcome matters to a decision, whether results apply to another setting, or why estimates differ.
The exercise uses fictional task-time values and a normal-approximation fixed-effect calculation solely to teach the graphic. It does not report a trial, product test, learner outcome or measured benefit. For a confidence-interval practice with separate constructed summaries, continue to the research-article reading worksheet.
Questions about this topic
What does the square in a forest plot mean?
It marks an individual study’s point estimate. In a meta-analysis forest plot, its area usually represents the study’s assigned weight. Read the legend and methods because visual conventions and models can vary.
What does a horizontal line through a study square show?
It usually represents that study estimate’s confidence interval. Compare it with the axis and null value, then keep the outcome, units, comparison direction and interval level in view.
What does the diamond at the bottom mean?
When present, the diamond usually shows the pooled estimate at its center and its confidence interval across its width. Not every forest plot includes a pooled result.
What is the null value in a forest plot?
It depends on the measure. A difference measure such as a mean difference has a null of zero; a ratio measure such as a risk ratio has a null of one.
Can a forest plot show results without a meta-analysis?
Yes. A plot can present individual study estimates when a pooled analysis is not possible or appropriate. The lack of a diamond means no pooled estimate is displayed; it does not by itself explain why.
Can I determine heterogeneity from confidence-interval overlap?
Poor overlap can be a visual clue, but it is not a complete heterogeneity assessment. Review the methods, formal measures, number of studies and possible clinical or methodological differences.
Sources and editorial notes
The supporting reading below informs our editorial guidance. Exercises and scenarios illustrate a method; they do not report a completed product test or measured learning outcome. Read our editorial policy or send a correction with a supporting source to info@ai-essay.com.