How to Interpret a Forest Plot (A Step-by-Step Guide)

A forest plot is the single most information-dense figure in a systematic review, and also one of the most frequently misread. Once you know what each element represents, it takes seconds to extract the same information that would otherwise require reading an entire results table.

The basic anatomy

Each horizontal line in a forest plot represents one included study. The position of the point estimate on that line -- usually a square or box -- shows the study's measured effect size, whether that is a risk ratio, odds ratio, mean difference, or another effect measure specified in your protocol. The horizontal line extending from that box represents the confidence interval, almost always the 95% confidence interval, showing the range of effect sizes consistent with that study's data.

The size of the box is not decorative. It typically represents the weight that study contributes to the pooled estimate, which is usually driven by its precision -- larger studies with narrower confidence intervals get bigger boxes and more influence on the final pooled result.

Reading the vertical reference line

A vertical line runs down the plot at the value representing "no effect" -- 1.0 for ratio measures like risk ratio or odds ratio, or 0 for difference measures like mean difference. Any study whose confidence interval crosses this line does not show a statistically significant effect at conventional thresholds, because the interval includes the possibility of no difference between groups.

Studies entirely to one side of the line, without their confidence interval touching it, show a statistically significant effect in that direction. Which side favors which group is specified in axis labels at the bottom of the plot, typically something like "Favors control" and "Favors intervention" -- always check this labeling explicitly, since the direction is not standardized across every plot and misreading it inverts your interpretation entirely.

The diamond at the bottom

The diamond shape at the bottom of the plot represents the pooled effect estimate across all included studies, with the diamond's width representing its confidence interval. This is the headline number of your meta-analysis: where the diamond sits relative to the line of no effect, and whether its width touches that line, tells you whether the pooled result is statistically significant.

The diamond's horizontal position is not simply an average of the individual study estimates -- it is a weighted average, where more precise (typically larger) studies pull the pooled estimate more strongly toward their own result, following whichever weighting scheme your fixed-effect or random-effects model uses.

Reading heterogeneity directly from the plot

Before you even look at the reported I-squared statistic, a forest plot gives you a visual sense of heterogeneity. If the individual study boxes cluster tightly around a similar value with substantially overlapping confidence intervals, heterogeneity is likely low. If the boxes are scattered across a wide range with confidence intervals that barely overlap or do not overlap at all, that visual spread is heterogeneity made visible, and it should prompt you to check the formal statistics before trusting the pooled estimate at face value.

A forest plot where the point estimates fall clearly on both sides of the line of no effect -- some studies favoring one group, others favoring the opposite group, both with reasonably narrow confidence intervals -- is a particularly strong visual signal of heterogeneity that a single pooled number may be obscuring rather than summarizing.

What the study order tells you

Studies in a forest plot are commonly ordered either chronologically by publication year or by weight, largest to smallest. Chronological ordering lets you visually assess whether effect sizes have shifted over time, which can be a signal of evolving practice, publication bias favoring early dramatic results, or changes in the underlying population being studied. Weight-based ordering makes it easier to see at a glance which studies are actually driving the pooled result.

Subgroup panels within a single plot

Many forest plots present subgroup analyses as stacked panels within the same figure, each with its own set of studies and its own subgroup diamond, followed by an overall diamond at the very bottom. When reading these, compare the subgroup diamonds to each other, not just to the line of no effect -- meaningfully different subgroup diamond positions suggest an effect modifier worth discussing, even if each subgroup individually shows a similar direction of effect.

A quick checklist for reading any forest plot

Identify the effect measure and confirm what value represents no effect. Check the axis labels to know which direction favors which group. Scan the individual study boxes for how tightly or loosely they cluster. Look at the bottom diamond for the pooled estimate and whether it crosses the no-effect line. Cross-reference the visual spread of studies against the reported heterogeneity statistics rather than assuming visual clustering by itself confirms low heterogeneity.

A forest plot rewards a slow first read. Once you know these six or seven elements, reading a new one takes seconds --

Spotting a study that does not belong

Occasionally a single study's box sits far from every other study on the plot, with a confidence interval that barely overlaps the rest of the cluster. This visual outlier is worth investigating rather than ignoring -- it may reflect a genuinely different population, a methodological limitation not fully captured by your risk-of-bias assessment, a difference in how the outcome was measured or defined across studies, or simply a very small sample producing an imprecise, unstable estimate that should not be dismissed without a closer look. A sensitivity analysis excluding that single study, reported alongside the full pooled estimate, tells the reader directly how much that one outlier is driving your overall conclusion.

A forest plot rewards a slow first read. Once you know these six or seven elements, reading a new one takes seconds -- and you will often catch inconsistencies between what the plot shows and what the text claims, which is exactly the kind of discrepancy a careful peer reviewer is trained to look for.