How to Read a Forest Plot: An Annotated Worked Example from a Meta-Analysis

If you have opened a systematic review and felt lost among the rows of boxes and lines, this guide on how to read a forest plot is for you. A forest plot is the signature graphic of a meta-analysis: it stacks the results of several studies on one chart and adds one combined, or pooled, result at the bottom [1]. Having a forest plot explained in plain language lets you see at a glance how consistent the evidence is and which way it points. Each study contributes one horizontal row, so a single chart summarises a whole body of research [4].

Before you interpret a forest plot, find three features. First, the vertical reference line is the line of no effect. For a ratio measure such as an odds ratio or a risk ratio it sits at 1.0; for a difference measure it sits at 0. Second, each square is a study’s point estimate, its best single-number result. The size of the square reflects the study weight, so larger, more precise studies draw a bigger box [1]. Third, the horizontal line through each square (its whiskers) is the 95% confidence interval. A narrow interval signals a precise estimate; a wide one signals uncertainty [3].

Work down the plot one row at a time and ask three questions of every study. Where does the square fall relative to the line of no effect? When the outcome is harmful, such as death or relapse, a square to the left of 1.0 favours the treatment and one to the right favours the control. Does the confidence interval cross the line of no effect? If it does, that study’s result is not statistically significant on its own. How wide is the interval? The answer tells you how much to trust the point estimate [4].

In Figure 1, Study D sits furthest left, Study C straddles the line, and Study F has the widest interval and the smallest square. Learning to interpret a forest plot is largely this habit of scanning each row against the reference line.

At the foot of the chart is the diamond, the pooled effect: a weighted average of all the studies. Its centre marks the summary estimate and its width marks the summary confidence interval, so you read the diamond just as you read a study, with more data behind it [1]. The guiding rule of any forest plot meta-analysis is simple. If the diamond touches or crosses the line of no effect, the pooled result is not statistically significant; if it sits clear of the line, it is. In this worked example the diamond is centred at an odds ratio of 0.81, with a confidence interval of 0.73 to 0.89 that sits clear of 1.0, so the combined evidence favours the treatment [3].

A pooled number only means something if the studies are reasonably alike. Heterogeneity describes how much the true effect varies from study to study. The most reported measure is I-squared (I²), the percentage of variation across studies that is due to real differences rather than chance [2]. Its authors proposed rough bands: about 25% is low, 50% moderate and 75% high [2]. Treat these as guides, not cut-offs, because I² is imprecise when only a few studies are pooled.

Visually, when the squares line up and the intervals overlap, heterogeneity is low; when squares scatter and whiskers barely touch, it is high. The six studies in Figure 1 give I² = 0% (Q = 3.44 on 5 degrees of freedom), so the variation is no more than chance would produce [2].

The table and the plot carry the same information in two forms. The study weight column explains why some squares are larger: Study E carries 22.5% of the weight and so pulls the pooled effect more than Study F at 7.8%. Weights here come from the inverse of each study’s variance, so the narrowest intervals earn the most weight. The “crosses line?” column mirrors what your eye does on the chart: any study whose interval spans 1.0 is individually non-significant, even when its point estimate suggests benefit [4]. Reading the two together is the fastest way to sanity-check a published forest plot before you trust its headline number. For effect sizes in more depth, see our forest plot interpretation guide.

This forest plot example tells a clear story. Four of six studies are individually non-significant, yet all lean towards benefit, and pooling them gives a precise, significant odds ratio of 0.81. Low heterogeneity means the studies agree. It does not prove they are unbiased, so the next step is to appraise risk of bias and publication bias. That is the whole skill of how to read a forest plot: from single rows, to the diamond, to a judgement about consistency.