You have had a meta-analysis performed and there is a forest plot in front of you. When the committee or a reviewer questions this graphic, what will you say? The forest plot is the most powerful tool for presenting the summary of a meta-analysis in a single visual; reading it correctly is critically important.
The Anatomy of a Forest Plot
Left column (Study list): Each included study is listed as a row, usually with the author's surname and year.
Center column (Visual): Contains a square and a horizontal line for each study.
- Square: Point estimate (effect size)
- Size of the square: The weight of the study (large sample = large square)
- Horizontal line: 95% confidence interval
Right column: The numerical value and weight (%) of each study
Bottom row (Diamond): The summary of the overall effect size.
- Center of the diamond: Overall effect size
- Width of the diamond: The 95% confidence interval of the overall estimate
The Vertical Line (Line of No Effect)
The vertical line passing through the center of the forest plot is very important:
- For Risk Ratio or Odds Ratio: This line is at 1
- For Mean Difference or Standardized Mean Difference: This line is at 0
If a study's confidence interval crosses this line, it means that study does not carry statistical significance individually.
Reading Heterogeneity from the Forest Plot
Squares spread over a wide area indicate high heterogeneity. The I² statistic shows this situation numerically:
- I² < 25%: Low heterogeneity
- I² between 25-75%: Moderate heterogeneity
- I² > 75%: High heterogeneity
In high heterogeneity, a random-effects model should be used instead of a fixed-effects model.
Frequently Asked Questions About Diamonds
"Does the diamond cross the vertical line?"
If it does, the overall effect is not statistically significant (p > 0.05).
"How wide is the diamond?"
Wide diamond = wide confidence interval = uncertain estimate. This usually occurs due to a small number of studies or a small sample size.
"Fixed vs random effects?"
If heterogeneity is low, a fixed-effects model is preferred; if it is high, a random-effects model is preferred. The results of both models can be compared and presented as a sensitivity analysis.
For forest plot interpretation and meta-analysis support, get a free consultation.
Where Do People Get Stuck Most in This Analysis?
- The I² value in the forest plot came out at 80%, and the question of whether the meta-analysis result is reliable in this case remains unanswered.
- The committee asks "why did you choose a random-effects model," but you cannot explain the rationale for the fixed vs random decision.
- You need to perform a subgroup analysis, but the subgroups are too small and the statistical power is insufficient.